CN112395885A - Short text semantic understanding template generation method, semantic understanding processing method and device - Google Patents
Short text semantic understanding template generation method, semantic understanding processing method and device Download PDFInfo
- Publication number
- CN112395885A CN112395885A CN202011359958.1A CN202011359958A CN112395885A CN 112395885 A CN112395885 A CN 112395885A CN 202011359958 A CN202011359958 A CN 202011359958A CN 112395885 A CN112395885 A CN 112395885A
- Authority
- CN
- China
- Prior art keywords
- preset
- semantic
- template
- item
- minimum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000003672 processing method Methods 0.000 title claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 40
- 230000004044 response Effects 0.000 claims description 86
- 238000007781 pre-processing Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 11
- 238000004891 communication Methods 0.000 claims description 5
- 230000002452 interceptive effect Effects 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 230000003993 interaction Effects 0.000 description 20
- 230000006870 function Effects 0.000 description 9
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 101001136140 Pinus strobus Putative oxygen-evolving enhancer protein 2 Proteins 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 210000001072 colon Anatomy 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 235000010799 Cucumis sativus var sativus Nutrition 0.000 description 1
- 244000299906 Cucumis sativus var. sativus Species 0.000 description 1
- 206010038743 Restlessness Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000002354 daily effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000001061 forehead Anatomy 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 231100000862 numbness Toxicity 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Machine Translation (AREA)
Abstract
A short text semantic understanding template generation method, a semantic understanding processing method and a device are provided, the method comprises the following steps: step S1, collecting a first preset number of short texts to be semantically understood; step S2 is to generate a preset basic semantic group set according to a first preset number of short texts to be semantically understood; step S3, generating and/or optimizing rule template sets based on the short texts to be semantically understood in the first preset number, and further generating and/or optimizing matching template sets corresponding to the preset basic semantic group sets; and step S4, generating a preset basic template set corresponding to the preset basic semantic set by combining the matching template set with the priority. The method comprises the steps of dividing the semantics contained in the short text to be semantically understood into mutually exclusive semantic groups, combining the priority and a matching template, and realizing the understanding processing of the short text to be semantically understood expressed by the natural language with human port phonetization, wherein the understanding processing comprises the processing of multiple intentions of the short text, the inconsistency of local semantics and overall semantics and the processing of part word errors caused by voice recognition.
Description
Technical Field
The application relates to the field of natural language semantic understanding, in particular to a short text semantic understanding template generation method, a semantic understanding processing method and a semantic understanding processing device.
Background
The human-computer intelligent interaction through the recognition, understanding and expression of the human natural language brings great convenience to the life of people, and along with the continuous improvement of the accuracy rate and the generalization capability of voice recognition, the human-computer intelligent interaction is gradually applied to various products and is familiar to the public. One of the key technologies affecting popularization and public experience is understanding of short text of natural language after speech recognition.
Short text understanding methods are broadly classified into matching template-based and machine learning model-based methods. The former is mostly keyword extraction, (after word segmentation) synonymy combination, abnf grammar. The existing method has complex flow, is difficult to manage redundancy, is difficult to balance accuracy and matching rate, and cannot process the classification of relatively complex semantics, such as inconsistent local semantics and overall semantics, human port linguisticization in an actual interactive scene instead of expression of strict written language, and the requirements of different semantic widths in the actual interactive scene. The method based on the machine learning model needs a large amount of actual scene interaction data, time is consumed for labeling, and professional research and development personnel do training to obtain a model reaching a certain accuracy rate, so that the starting cost is high, and small or non-professional enterprises are difficult to use, and further no way of generating the actual scene interaction data is caused.
Disclosure of Invention
The application provides a short text semantic understanding template generation method, a semantic understanding processing method and a semantic understanding processing device, and aims to solve the problems that short text semantic understanding accuracy is not high enough, service fusion is difficult, and actual production and use are not supported enough in the prior art under the condition of lack of actual production data.
In order to achieve the above object, the present application provides the following technical solutions:
a short text semantic understanding template generation method comprises the following steps:
step S1: collecting a first preset number of short texts to be semantically understood;
step S2: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set consists of a plurality of preset basic semantic groups which do not have inclusion or included relationship;
step S3: generating and/or optimizing a rule template set based on a first preset number of short texts to be semantically understood, and further generating and/or optimizing a matching template set corresponding to a preset basic semantic group set, wherein the rule template comprises a plurality of regular template components and is a minimum unit for performing short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the matching templates are divided into a formal class and an auxiliary class, and each class consists of a plurality of corresponding rule templates;
step S4: and the matching template set is combined with the priority to generate a preset basic template set corresponding to the preset basic semantic set.
In the foregoing solution, the step S2 generates a preset basic semantic group set according to a first preset number of short texts to be semantically understood, including:
step S21: generating a semantic library, wherein the semantic library consists of all nonrepeating minimum semantic units in a first preset number of short texts to be semantically understood; the minimum semantic unit refers to the semantics of a minimum number of clause combinations; the clauses refer to Chinese punctuations connected with text contents and then Chinese punctuations; the minimum number of clause combinations means that the semantics of the clause combinations smaller than the minimum number in the natural language expression cannot be understood and therefore cannot be reasonably responded, and the semantics of the clause combinations reaching the corresponding minimum number can be understood and responded;
step S22: and classifying the minimum semantic unit in the semantic library into a plurality of preset basic semantic groups with preset semantic widths according to semantic classification corresponding to a response range preset by the service or the field, and generating a preset basic semantic group set.
In the foregoing solution, the step S3 further generates and/or optimizes a matching template set corresponding to a preset basic semantic group set based on a first preset number of short text to be semantically understood and/or an optimization rule template set, including:
step S31: selecting a third preset number of short texts to be semantically understood from the first preset number of short texts to be semantically understood, wherein the third preset number of short texts to be semantically understood simultaneously meets 2 conditions of related minimum semantic units S1, and the minimum semantic units S1 are classified in a preset basic semantic group BSi;
the 2 conditions are:
the condition is that the short text to be semantically understood contains the semantics of the minimum semantic unit S1;
the second condition is that the semantic of the minimum semantic unit S1 is consistent with the overall semantic of the whole short text to be semantically understood;
step S32: extracting key characters in a plurality of minimum length combinations for determining the semantics of a minimum semantic unit S1 from a third preset number of short texts to be semantically understood, wherein the length refers to the number of the key characters; the minimum length combination is divided into a formal class and an auxiliary class according to whether ambiguity is generated during understanding;
step S33: selecting a proper regular template component according to the types of the key characters in each extracted minimum length combination, and generating and/or optimizing a rule template set corresponding to a minimum semantic unit S1 by combining the relative positions of the key characters, wherein the types of the key characters comprise synonym characters, forbidden characters, synonym out-of-order n-time common characters, and 4 types of special characters which are partially or completely forbidden;
step S34: according to the type of each minimum length combination, putting the corresponding rule template into the corresponding formal class or auxiliary class set to form a set of preset matching templates corresponding to the minimum semantic unit S1;
step S35: sequentially repeating the steps for other minimum semantic units contained in the preset basic semantic group BSi to generate a preset matching template set of the preset basic semantic group BSi;
step S36: and circulating the steps to generate a preset matching template set of all preset basic semantic groups in the preset basic semantic group set.
In the above scheme, the step S33 selects a suitable regular template component according to the extracted category of the key character in each minimum length combination, and generates and/or optimizes a rule template set corresponding to the minimum semantic unit S1 by combining the relative positions of the key characters, including:
step S331: preprocessing one short text to be semantically understood in a third preset number of short texts to be semantically understood;
step S332: finding out a minimum number of clause combinations which accord with the semantics of the whole short text and contain key characters in the found minimum length combinations from the preprocessed short text to be semantically understood;
step S333: selecting a corresponding regular template component according to the key character type in each minimum length combination in the minimum number of clause combinations;
step S334: combining the relative positions of the key characters in the minimum number of clause combinations found out and the use positions of the selected regular template components to generate or optimize a preset rule template;
step S335: repeating the above processes for a third preset number of short texts to be semantically understood, generating and/or optimizing a plurality of rule templates corresponding to the minimum semantic unit S1 in each short text, and combining the rule templates into a rule template set corresponding to the minimum semantic unit S1.
In the foregoing solution, the step S334, in combination with the found relative position of the key character in the minimum number of clause combinations, and the selected use position of the regular template component, generates or optimizes the preset rule template, including:
searching whether a preset rule template exists in a formal template set and an auxiliary template set in a preset basic matching template corresponding to a corresponding preset basic semantic group to meet the requirement of a selected regular template component and a position connection structure thereof, wherein the position connection structure refers to the relative position of each part of the rule template determined by the relative position of a key character and the use position of the regular template component;
if the applicable preset rule template exists, the key characters are filled in the corresponding positions, and the applicable preset rule template is optimized, so that the preset rule template can be matched with the preprocessed short text to be semantically understood;
and if no applicable preset rule template exists, filling the key characters into the corresponding regular template components, and connecting and generating the applicable preset rule template according to the relative positions of the key characters and the using positions of the components, so that the preset rule template can be matched with the preprocessed short text to be semantically understood.
In the above solution, the step S34, according to the type of each minimum length combination, puts the corresponding rule template into the corresponding formal class or auxiliary class set, and forms a set of preset matching templates corresponding to the minimum semantic unit S1, including:
step S341: putting a rule template formed by key characters in the formal class minimum length combination into a corresponding formal class template set;
step S342: putting a rule template formed by key characters in the auxiliary class minimum length combination into a corresponding auxiliary class template set;
step S343: and combining the formal template set and the auxiliary template set to form a preset matching template set corresponding to the minimum semantic unit S1.
In the foregoing solution, the step S4 of generating, by combining the matching template set with the priority, a preset basic template set corresponding to the preset basic semantic set includes:
step S41: dividing all preset basic semantic groups into a plurality of large categories according to response content, modes and purposes required by services or fields, determining the priority among the large categories, and generating a preset layer 1 priority;
step S42: determining the priority among preset basic semantic groups in each large category according to an actual reasonable response sequence, and generating a preset layer 2 priority;
step S43: and combining the matching template set corresponding to each preset basic semantic group in the preset basic semantic group set with the 2-layer priority to form a preset basic template group set.
The application also provides a short text semantic understanding processing method, which comprises the following steps:
step M1: generating each preset item semantic group set and a corresponding preset item template group set;
step M2: acquiring a short text to be semantically understood;
step M3: preprocessing short texts to be semantically understood;
step M4: matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching;
step M5: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode.
In the foregoing solution, the step M1 of generating each preset item semantic group set and a corresponding preset item template group set includes:
step M11: classifying the minimum semantic unit in the semantic library into a plurality of preset item semantic groups with preset semantic widths according to semantic classification corresponding to the preset response range of each item, and generating a preset item semantic group set corresponding to each item;
step M12: generating one-to-one or one-to-many corresponding relation between each preset item semantic group and a preset basic semantic group in the preset item semantic group set of the corresponding item by combining the preset basic semantic group set according to the preset item semantic group set of each item;
step M13: and combining the priority, the corresponding relation between each preset item semantic group in each preset item semantic group set and each preset basic semantic group in the preset basic semantic group set, presetting the preset basic module group set corresponding to the basic semantic group set, and generating the preset item module group set corresponding to each preset item semantic group set.
In the above scheme, the method further comprises:
in a corresponding short text semantic understanding processing link in a human-end and machine-end interactive scene, performing short text semantic understanding processing by using a corresponding preset item semantic group set and a corresponding preset item template group set, and collecting short texts to be semantically understood which are not repeated with the collected short texts; and optimizing the generated preset item semantic group set by using the collected and collected nonrepeated short texts to be semantically understood in combination with the continuously adjusted business requirements, and further optimizing the preset basic semantic group set, the preset basic template group set and each preset item template group set.
The present application further provides a semantic understanding template generating apparatus, including:
a short text collection unit: collecting a first preset number of short texts to be semantically understood;
a basic semantic group generating unit: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set consists of a plurality of preset basic semantic groups which do not have inclusion or included relationship;
a template generation unit: generating and/or optimizing a rule template set based on a first preset number of short texts to be semantically understood, and further generating and/or optimizing a matching template set corresponding to a preset basic semantic group set, wherein the rule template comprises a plurality of regular template components and is a minimum unit for performing short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the matching templates are divided into a formal class and an auxiliary class, and each class consists of a plurality of corresponding rule templates;
a basic template group set generation unit: and the matching template set is combined with the priority to generate a preset basic template set corresponding to the preset basic semantic set.
The present application also provides a semantic understanding processing apparatus, including:
a project template group generating unit: generating each preset item semantic group set and a corresponding preset item template group set;
a short text acquisition unit: acquiring a short text to be semantically understood;
a short text preprocessing unit: preprocessing short texts to be semantically understood;
a matching unit: matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching;
a response unit: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode.
The application also provides an electronic device, which comprises a processor, a memory, a computer program stored on the memory and capable of running on the processor, a communication bus and short text semantic understanding processing interfaces of all items, wherein the processor realizes the semantic understanding template generation method and the semantic understanding processing method when executing the program.
The application also provides a computer readable storage medium, on which executable instructions are stored, and the executable instructions are executed by a processor to realize the semantic understanding template generation method and the semantic understanding processing method.
Compared with the prior art, the beneficial effects of this application are:
the method has the advantages that semantic groups with various thickness semantic widths are preset, a template group (including priority) corresponding to the semantic group is generated by combining a designed rule template generation flow, a response mode of a project is further combined, the requirement for semantic understanding of natural language of human port linguisticization in man-machine interaction service can be met flexibly, particularly, the problem of semantic understanding complexity caused by local semantics and integral semantics inconsistency, long short text multiple semantics, short text, rule template redundancy and individual character errors after voice recognition can be solved, and the response of controllable categories after semantic understanding is realized; by setting a preset basic semantic group and a preset project semantic group, a template group set corresponding to a set of basic semantic group set can be used for a plurality of projects, the actual service requirement is met, and the production efficiency is improved; in addition, the method supports semantic understanding processing when a small amount of data exists and response of a machine end during man-machine interaction, and can continuously improve the accuracy of natural language semantic understanding processing in interaction along with the increase of data quantity.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a short text semantic understanding template generation method disclosed in the embodiment of the present application;
FIG. 2 is a flow chart of basic semantic group generation in the short text semantic understanding template generation method disclosed in the embodiment of the present application;
fig. 3 is a flowchart of generating a matching template corresponding to a basic semantic group in the short text semantic understanding template generating method disclosed in the embodiment of the present application;
fig. 4 is a flowchart of generating a rule template set corresponding to a single minimum semantic unit in the short text semantic understanding template generating method disclosed in the embodiment of the present application;
fig. 5 is a flowchart of generating a matching template corresponding to a single minimum semantic unit in the short text semantic understanding template generating method disclosed in the embodiment of the present application;
fig. 6 is a flowchart of generating a basic template set in the short text semantic understanding template generating method disclosed in the embodiment of the present application;
FIG. 7 is a flowchart of a short text semantic understanding processing method disclosed in an embodiment of the present application;
fig. 8 is a flowchart of generating a project template set in the short text semantic understanding processing method disclosed in the embodiment of the present application;
fig. 9 is a schematic structural diagram of a short text semantic understanding template generating apparatus disclosed in the embodiment of the present application;
fig. 10 is a schematic structural diagram of a short text semantic understanding processing apparatus disclosed in an embodiment of the present application;
fig. 11 is a schematic structural diagram of an apparatus disclosed in an embodiment of the present application.
Detailed Description
The embodiment of the application discloses a short text semantic understanding template generation method, a semantic understanding processing method and a semantic understanding processing device, which can be applied to electronic equipment with man-machine intelligent interaction, such as intelligent dialogue equipment, intelligent home equipment, intelligent wearing equipment and the like.
The technical solution in the embodiment of the present application will be clearly and completely described below with reference to the drawings in the embodiment of the present application and an intelligent hastening robot system as an example, and it is obvious that the described embodiment is a part of the embodiment of the present application, but not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Fig. 1 is a method for generating a short text semantic understanding template disclosed in the embodiment of the present application, including the following steps:
step S1: a first preset number of short texts to be semantically understood is collected. Specifically, according to a response range preset by a service or a field, a first preset number of pieces of short texts to be semantically understood after duplication removal can be designed or extracted and stored in a short text library to be semantically understood.
The preset response range of the service or the field refers to the sum of the preset response ranges of all the projects; the project refers to a project aiming at a specific service, such as intelligent conversation, intelligent home, intelligent wearing and the like which are directly oriented to terminal clients by each manufacturer; the preset response range of a certain item refers to the sum of corresponding response contents made by the machine after understanding the semantics of various natural languages expressed by the human end, for example, in a certain collection item, the preset unrepeated response contents have a second preset number of pieces, including a plurality of pieces of bottom-of-pocket response contents; the bottom-in response content refers to a response which is given by the machine and accords with various conditions under the condition that other response contents are not suitable, for example, "I do not understand the meaning of your expression and ask for follow-up consultation customer service. ".
For convenience of description, the short text to be semantically understood mentioned below refers to a piece of text from the human end and expressing contents in natural language in one round of interaction in the intelligent interaction process between the human end and the machine end, wherein if the contents expressed in natural language belong to a speech form, the contents in the speech form expressed in natural language need to be converted into corresponding contents in the text form expressed in natural language with punctuation marks through a speech recognition engine.
The punctuation mark in the short text to be semantically understood has the function of enabling the rule template to process each semantic in the short text to be semantically understood in the form of text. Short pauses in the content speech form expressed in natural language and punctuation marks in the text form play an important role in treating each local semantic and overall semantic of the semantically understood short text. The rule template comprises a plurality of regular template components which are the minimum units for carrying out short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the punctuation combination is the combination of punctuation marks in a short text to be semantically understood, and the action is equivalent to the action of short pause in a content voice form expressed by a natural language and the punctuation marks in a text form on the aspect of semanteme understanding; the key characters refer to Chinese characters (such as ' still '), Chinese character phrases (such as ' processing '), Chinese and English punctuations (such as ' 6: 30: ', Chinese comma '), English words or letters (such as ' APP, a, p ').
For example, "i have paid back in the background of the credit card, i do not know how much the cost of you is, i are ready to find that you help i inquire about how much the cost is, and i need to pay back every month. "i don't deal with, why you debt me", "i have already handed over a few days before, i do not know how your system is not updated, and still, and no consumption is done every month. "if Chinese punctuation is not added, there are places where a formal understanding of the text of a word can be ambiguous.
If there is no pause in the speech form expression of each short text to be semantically understood, or there is no punctuation in the text form expression, the content in the text form expressed in natural language without punctuation can also be used. Such as "how much I overdue", "why my card amount was dropped", "how much the card number.
The short text to be semantically understood of the first preset number after the duplication removal is designed or extracted refers to a question, an answer, an emotion expression and the like which can be artificially designed and generated or extracted from the actual interaction between the existing service seat and the customer.
Wherein the short text to be semantically understood extracted from the actual interaction contains wrong words, which relatively conforms to the actual scene.
The error word refers to an error word in the following two cases, namely, in the case of a short text to be semantically understood in the form of a text, the error word may contain a speech-like error word; in case of a short text to be semantically understood in the form of speech, the short text in the form of text recognized by the speech recognition engine may contain a word recognized by the speech recognition engine in error.
The deduplication refers to, for example, extracting or designing ten thousand short texts to be semantically understood, and each ten thousand short texts are required to be different.
Step S2: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set is composed of a plurality of preset basic semantic groups which do not have inclusion or included relationship with each other.
The detailed procedure of step S2 is described below in fig. 2.
Step S3: generating and/or optimizing a rule template set according to the collected short texts to be semantically understood in the first preset number, the preset basic semantic group set and the matching template generation flow chart, and further generating and/or optimizing a matching template set corresponding to the preset basic semantic group set.
The detailed procedure of step S3 is described below in fig. 3.
Step S4: and the matching template set is combined with the priority generation flow chart to generate a preset basic template set corresponding to the preset basic semantic set.
The detailed procedure of step S4 is described below in fig. 6.
Fig. 2 is a flow chart of generating a basic semantic group in a short text semantic understanding template generating method disclosed in the embodiment of the present application, including the following steps:
step S21: and generating a semantic library, wherein the semantic library consists of all nonrepeating minimum semantic units in a first preset number of short texts to be semantically understood.
The minimum semantic unit refers to the semantics of a minimum number of clause combinations; the clauses refer to Chinese punctuations connected with text contents and then Chinese punctuations; the minimum number of clause combinations means that semantics of less than the minimum number of clause combinations in the natural language expression cannot be understood and thus cannot be reasonably responded to, and semantics of up to the corresponding minimum number of clause combinations can be understood and responded to.
The minimum semantic units have no inclusion or included relationship with each other.
Step S22: classifying the minimum semantic units in the semantic library into a plurality of preset basic semantic groups with preset semantic widths according to semantic classification corresponding to a response range preset by a service or a field, and generating a preset basic semantic group set, so that the minimum semantic units related in a first preset number of short texts to be semantically understood can be classified into all preset basic semantic groups, and the same minimum semantic unit cannot be classified into a preset basic semantic group BSi (i is not equal to j) any more, and all the preset basic semantic groups do not have inclusion and included relations.
The semantic classification corresponding to the response range preset by the service or the field refers to a classification determined by the semantic classification corresponding to the response range preset by each item, for example, one of the preset semantic classifications of the item P1 is a combination of minimum semantic units S1, S2 and S3 all using the same response content preset by P1 as a response, and one of the preset semantic classifications of the item P2 is a combination of minimum semantic units S2, S3 and S4 all using the same response content preset by P2 as a response. The semantic classification corresponding to the preset response range of the business or the field needs at least 3 types, one type of S1, one type of S2 and S3, and one type of S4.
For example, bank a requires to urge m0 business, and the response contents corresponding to the minimum semantic units "good", "can", "today's debt" are all "that does not disturb you, and see again. "; the bank B requires to urge to collect the m0 business, the corresponding response contents of the minimum semantic units of 'can', 'debt is still in the day' and 'debt is still in the tomorrow' are 'good', and you can pay for payment through WeChat or Paibao and see again. "; here, semantic classification corresponding to a response range preset in a business or a field needs at least 3 types, namely, a "good" type, a "ok" type, a "today debt" combination type, and a "tomorrow debt" type.
The semantic width refers to each semantic classification corresponding to a response range preset by a service, a field or each item, for example, the combination/number of more than or equal to one minimum semantic unit corresponding to a specific response content. For example, a certain bank project requires that the preset response range of the m0 service be 5 replies, wherein the response contents corresponding to the minimum semantic units "good", "debt today" and "debt tomorrow" expressed by the human end are all "that does not disturb you, and see again. The sentence (1 in 5 replies) is replied, the preset response range is 5 replies, and various minimum semantics expressed by the human end are classified and respectively correspond to one of the 5 replies; here a predefined semantic width is associated with the response content "that does not disturb you, see again. The combination/number of "corresponding minimum semantic units" good "," debt still today "," debt still tomorrow ".
The width of the preset basic semantic group can be adjusted according to project requirements; in the narrowest case, each minimum semantic unit corresponds to a preset basic semantic group.
The requirements for understanding the natural language semantics of the human port linguisticization in the actual service can be flexibly met by presetting the preset basic semantic group with various thickness semantic widths.
Fig. 3 is a flowchart of generating a matching template corresponding to a basic semantic group in a short text semantic understanding template generating method disclosed in the embodiment of the present application, including the following steps:
step S31: and randomly selecting a preset basic semantic group BSi, supposing that the preset basic semantic group BSi comprises a minimum semantic unit S1, a minimum semantic unit S2 and a minimum semantic unit S3, and selecting a third preset number of short texts to be semantically understood from a short text library to be semantically understood, wherein the short texts to be semantically understood simultaneously meet 2 conditions of the minimum semantic unit S1. The 2 conditions are specifically as follows:
the condition is that the short text T1 to be semantically understood contains the semantics of the smallest semantic unit S1.
The second condition is that the semantics of the minimum semantic unit S1 are consistent with the overall semantics of the whole short text T1 to be semantically understood.
If a new matching template is generated for the first time, the larger the third preset number is, the better the third preset number is.
If the existing matching template is optimized, the third preset number can be determined according to the error rate, a small amount of optimization can be performed on the condition of low error rate, and the larger the third preset number is, the higher the optimization efficiency is, and the larger the error rate reduction amplitude is.
Step S32: and extracting key characters in a plurality of minimum length combinations for determining the semantics of the minimum semantic unit S1 from a third preset number of short texts to be semantically understood, which contain the semantics of the minimum semantic unit S1.
The length in the minimum length combination refers to the number of key characters used.
For example, the minimum semantic unit S1 is "the customer states that the customer has paid too much to receive the seat, and the short text to be semantically understood in a certain sentence pattern is as follows:
"i has not yet passed. ";
"i don't have a prayer wheel. ";
"i did not play with the prayer wheels. "(" play "word, true voice expressed" still ", caused by a speech recognition engine recognition error);
"I just returned. ";
"I still owes money. ";
"also go. ";
"go to bar. ";
……
representative key characters are: not, but, has, and is. The rest key characters have similar semantic functions with the above key characters, for example, similar to "already", there are "just", "yesterday" … …; similar to "also" are "save", "transfer", "play", "handle", "do", …, "change (caused by a speech recognition engine recognition error)", "yellow (caused by a speech recognition engine recognition error)", "play (caused by a speech recognition engine recognition error)" …
The minimum semantic unit S1- "the customer expresses that the customer has paid too late for the seat to be charged," the various minimum length combinations of the sentence pattern in the example are shown in table 1, each row represents a length combination, and the column element corresponding to the opposite check in each row is the representative key character in the length combination represented by the row.
TABLE 1
Is not limited to | Is that | Has already been used for | And also | For treating | To master | Ma | |
1 | √ | √ | |||||
2 | √ | √ | |||||
3 | √ | √ | √ | ||||
4 | √ | √ | √ | √ | |||
5 | √ | √ |
The minimum length combinations are classified into a main class and an auxiliary class according to whether ambiguity occurs when understanding. The formal class refers to minimum length combinations that do not generate ambiguity, such as 4 minimum length combinations represented by lines 1 to 4 in table 1; the auxiliary class refers to the minimum length combination that will produce ambiguity, such as the minimum length combination represented by row 5 in table 1.
Step S33: and selecting a proper regular template component according to the type of the key characters in each minimum length combination extracted by combining the rule template generation flow chart, and generating and/or optimizing a rule template set corresponding to the minimum semantic unit S1 according to the relative positions of the key characters, wherein the key characters comprise 4 categories of synonymous characters, forbidden characters, synonymous disordered n-order common characters and partially or completely forbidden special characters.
The detailed procedure of step S33 is shown in fig. 4 below.
Step S34: and according to the category of each minimum length combination, putting the corresponding rule template into the corresponding formal class or auxiliary class set to form a set of preset matching templates corresponding to the minimum semantic unit S1.
The detailed procedure of step S34 is shown in fig. 5 below.
Step S35: and repeating the steps for the minimum semantic unit S2 and the minimum semantic unit S3 to generate a preset matching template set of the preset basic semantic group BSi.
Step S36: and repeating the steps for other semantic groups in the preset basic semantic group set in sequence to generate a preset matching template set of the preset basic semantic group set.
Fig. 4 is a flowchart for generating a rule template set corresponding to a single minimum semantic unit in a short text semantic understanding template generating method disclosed in the embodiment of the present application, where the flowchart specifically includes the following steps:
step S331: and preprocessing one short text to be semantically understood in a third preset number of short texts to be semantically understood, specifically adding characters which identify the head end and the tail end of the short text in a designed rule template to the head end and the tail end of the short text to be semantically understood respectively. The characters refer to preset punctuation marks or English letters or words. The preprocessing function is to enable the rule template to identify the sentence beginning and sentence end identification of the preprocessed short text to be semantically understood, so as to distinguish the sentence beginning and sentence end identification from other parts of the short text to be semantically understood, and particularly distinguish the head and tail clauses and other clauses in the short text to be semantically understood. When the clause is positioned at the head end (namely the first clause) of the preprocessed short text to be semantically understood, characters for marking the head end exist at the initial position; when the clause is positioned at the tail end (namely, the tail clause), the tail end position is provided with a character for marking the tail end; when the clause is not a head-to-tail clause, the starting position and the tail position have only one Chinese punctuation.
For example, the beginning is denoted as "START" and the END is denoted as "END".
For another example, two Chinese colon symbols are added at the head end and one Chinese period symbol is added at the tail end, which is suitable for preprocessing short text to be semantically understood in the form of text expressed in natural language with punctuation marks.
For another example, two Chinese colon symbols are added at the head end and two Chinese period symbols are added at the tail end, which is suitable for preprocessing short text to be semantically understood in the form of text expressed in natural language with punctuation marks and for preprocessing short text to be semantically understood in the form of text expressed in natural language without punctuation marks. In the present embodiment, such a flag is used for explanation, and other characters with similar functions are also within the protection scope.
For example, the short text to be semantically understood is: how do you, I have also. ", the pretreatment is: ": : how do you, I have also. . . ".
Step S332: and finding out a minimum number of clause combinations which accord with the semantics of the whole short text and contain key characters in the above found multiple minimum length combinations from the preprocessed short text to be semantically understood.
Step S333: and selecting a corresponding regular template component according to the key character type in each minimum length combination in the found minimum number of clause combinations.
The specification of the regularization, punctuation combination, key characters, usage location and function in each component of the regularization template is given in table 2 below.
TABLE 2
The regularization in table 2 may be substituted with other representations to achieve equivalent functionality.
The punctuation combinations in table 2 may be replaced with other representations that achieve equivalent functionality, the present application with punctuation returned by the speech recognition engine used. | A Is there a "plus pretreatment added at head end": "for example, the remaining equivalent functions are also within the scope of protection.
Step S334: and combining the relative positions of the key characters in the minimum number of clause combinations found out and the use positions of the selected regular template components to generate or optimize a preset rule template.
Searching whether a preset rule template exists in a formal template set and an auxiliary template set in a preset basic matching template corresponding to the corresponding preset basic semantic group, and meeting the requirements of the regular template component selected in the step 334 and a position connection structure thereof, wherein the position connection structure refers to the relative position of each part of the rule template determined by the relative position of the key character and the use position of the regular template component.
If the applicable preset rule template exists, the key characters are filled in the corresponding positions, and the applicable preset rule template is optimized, so that the preset rule template can be matched with the preprocessed short text to be semantically understood; and if no applicable preset rule template exists, filling the key characters into the corresponding regular template components, and connecting and generating the applicable preset rule template according to the relative positions of the key characters and the using positions of the components, so that the preset rule template can be matched with the preprocessed short text to be semantically understood.
Step S335: repeating the above processes for a third preset number of short texts to be semantically understood, generating and/or optimizing a plurality of rule templates corresponding to the minimum semantic unit S1 in each short text, and combining the rule templates into a rule template set corresponding to the minimum semantic unit S1.
For each short text to be semantically understood, the number of semantic groups to be matched may be one or more, and the specific number is determined according to the number of minimum semantic units contained in the short text itself and the number of semantic groups distributed in the project request response, for example, "three months are still clear, and this month is a few numbers from the beginning? "(here, the voice" up "is recognized as" three "by the voice recognition engine), a particular item may respond to both semantic groups" has been cleared the last month "and" which day the payment date is ", and then the short text should match the rule templates of both semantic groups.
Short texts to be semantically understood in the following examples are all short texts in a text form, which are recognized by a speech recognition engine in actual human-computer interaction, wherein short texts in a speech form expressed by a human end have partial recognition errors, and the examples are applied to illustrate the generation process and advantages of rule templates in the semantic understanding processing of the common nine-major short texts.
The category I is the situation that a rule template is newly added in a single meaning group, for example, the condition that the user does not help me to recover the quota, and the user does not remain. ".
Step S331: after pretreatment, the method comprises the following steps: : you do not help me to recover the amount, i do not. . . ";
step S332: the minimum number of clause combinations which are found to be consistent with the semantics of the whole short text and contain the key characters in the plurality of minimum length combinations found in the previous step are' as follows: : you do not help me to recover the amount, i do not. . . ";
step S333: the extracted key characters 'recovery' and 'forehead' belong to 2-class synonym characters, and a corresponding regular template component is selected ((;
And in the category II, the monolingual group optimizes the condition of the existing rule template, such as' you help me increase the quota first. "help me promote a little amount", "you give me some amount".
Optimizing existing preset rule templatesNamely, equivalent synonym characters are added (such as 'lifting' and 'adjusting' are equivalent to 'recovering'), regular template components are added (such as:
the equivalent synonymous characters and the regular template components added in the above example can reduce the redundancy of the rule template, and in the explanation of other categories, it can be seen that the added regular template components and their components can also reduce the redundancy of the rule template.
Category three, short text case of multiple semantic groups, such as "three months i have been clear yet, the month is a few numbers started yet? ",
formal class rule template of basic semantic group BSi in matching
Formal class rule template of basic semantic group BSj (j ≠ i) in matching
The long and short text multi-semantic matching processing is realized by reasonably selecting the designed regular template component, and the understanding processing of the long and short text multi-semantic can be realized by combining the preset semantic group in the overall design.
Category four, very short text case, such as: "which? "," parent ", that. "," take one, then? "the optimized rule template is:
by characterizing the punctuation combination of the first (":: |, |.
The fifth category, the case of character folding, is mostly used for very short texts, but other texts can be used as required.
"you say. "
"you say the bar, you say. "
"you say, feed, you say. "
Optimizing into the following rule templates:
class six, use of special characters which have been partially or totally banned, e.g.
"why do you not send me a short message? "
' the message sent to me just not just can be sent to me. "
The above 2 short texts should not match the rule template as follows, so the forbidden character "not" is added in the rule template,
in order to make the following short text "you can't send you a message to me do you see"? How did that be consumed? "in the matching, as the rule above, special characters and corresponding regular components that have been partially or completely forbidden are added" (.
And the seventh category is that the local clause semantics is inconsistent with the overall short text semantics, for example, so, I speak with your cheers, I need wait for two days before turning to, and I play. "(the actual speech at the human end is ' still ', and is recognized as ' playing ' after passing through a speech recognition engine), in the short text, the local clause ' i plays in. The semantic of the ' i just go in ' is inconsistent with the overall semantic of the short text, and the following rule template (with the addition of forbidden characters ' and the like) is adopted, so that the local and overall inconsistent semantic in the matching is avoided, and the short text of the ' i just go in ' with the local and overall semantic in the matching can be matched.
"I have returned, I have also wrongly blocked. ", local clause" i still. The semantics of the short text are inconsistent with the overall semantics of the short text, and the following rule template (with the addition of the prohibited character being wrong) is adopted, so that the local and overall inconsistent semantics in the matching process can not be matched, and the local and overall semantic consistency in the matching process can be matched, i still return, i return, and the like.
And the category eight is the condition that the local semantics is inconsistent with the overall single clause semantics.
The problem easily encountered by the conventional common synonym regular or wildcard is that the synonym template can match the clauses of positive and negative semantics, such as:
synonym regularization: also | also La | also past; wildcard character: and also
Can match "i have still", "i have still done cheer", "i have already done still", etc., but can also match "i have not yet", "i have forgotten to do still", "do i have not yet done yet? "
The rule templates (e.g., auxiliary class rule templates, (.
Category nine, in the case of speech recognition errors, e.g.,
"cucumber cheering". "(the actual human-end expressed speech is" still too cheer ").
"I am restless and cheerful. "(the actual human-end expressed speech is" I am cheer still ").
"the cushion has been replaced. "(the actual human-end expressed speech is" yesterday has been already ").
The auxiliary class rule templates of the corresponding basic semantic group in the matching are:
the similar error words are used as semantically equivalent keywords (such as 'vexation' and 'still' equivalent), and forbidden words for forbidding self semantemes (such as a method for forbidding a trouble to be a common expression of 'vexation' and limiting the self semantemes of 'vexation' by forbidding 'numbness') are added, so that the error words can be reasonably processed, the interaction difficulty caused by a client input error or a voice recognition engine recognition error is made up, and the whole interaction process is smoother.
In conclusion, through the rule template generation process, the problems of semantic understanding complexity caused by local semantics and overall semantics in actual interaction, long short text multiple semantics, short text, rule template redundancy and individual character errors after voice recognition can be solved.
Fig. 5 is a flowchart of generating a matching template corresponding to a single minimum semantic unit in the method for generating a semantic understanding template of a short text disclosed in the embodiment of the present application, including the following steps:
step S341: putting a rule template formed by key characters in the formal class minimum length combination into a corresponding formal class template set;
step S342: and putting the rule template formed by the key characters in the minimum length combination of the auxiliary class into the corresponding auxiliary class template set.
The formal template has higher accuracy, and the auxiliary template has higher matching success probability.
When the accuracy of the newly added rule template is not known, the rule template can be temporarily placed in the auxiliary class, the number of short texts of the rule template is optimized to be increased to a preset number, and the rule template is transferred to the formal class when the accuracy of the rule template is determined to reach a set threshold.
Step S343: and combining the formal template set and the auxiliary template set to form a preset matching template set corresponding to the minimum semantic unit S1.
The formal template set with high accuracy and the auxiliary template set with high matching probability are set, and the formal template set and the auxiliary template set are matched to realize high accuracy of natural language semantic understanding processing.
Fig. 6 is a flowchart for generating a basic template group in the short text semantic understanding template generation method disclosed in the embodiment of the present application, including the following steps:
assuming that the same short text to be semantically understood contains a plurality of preset basic semantic groups, a certain item is only suitable for responding to one of the preset basic semantic groups during one round of interaction or hopes to carry out combined response according to the sequence, and the method can be realized by setting the relative priority of each preset basic semantic group in the preset basic semantic group set.
Specifically, the multi-level priority may be set according to the service or domain requirements, for example, a layer is set between domains, a layer is set between specific large categories in the domains, and a layer is set between each preset semantic group in a specific large category in the domains. The following description will take two layers of priority as an example.
Step S41: dividing all the preset basic semantic groups into a plurality of large categories according to the response content, mode and use required by the service or field, determining the priority among the large categories, and generating the preset layer 1 priority.
For example, in the home field, the operation class with extremely short response has higher priority than the mood-soothing class.
"the room is hot and stuffy, the air conditioner is opened", in the current dialogue, the machine end is only suitable for responding to the situation of one type of semantic group, the air conditioner can be opened preferentially and the user can reply that the air conditioner is opened; when the machine side is suitable for the combined response, the combined response can be turned on according to the priority from high to low, and the user can reply that the air conditioner is turned on and the temperature in the room is gradually cooled down to Ha | within five minutes! ".
For example, in the field of collection, the priority for appeasing complaints is higher than that for answering questions, and the priority for continuing collection of debts is higher than that for continuing collection.
"why you did not have any trouble making a call, i complain you about your bank. "the machine side is only suitable for responding to one type, and can return the mood with priority" if our service brings inconvenience to you, please forgive the understanding! "; in the case of a machine-side adapted combined response, the combined response can be from high to low in priority, such as "if our service brings inconvenience to you, please forgive the understanding! Calling you every day is worried about more interest expenditure caused by delayed repayment and requires to repay the money according to the convention in time. ".
Step S42: and determining the priority among preset basic semantic groups in each large category according to an actual reasonable response sequence, and generating a preset layer 2 priority.
For example, in the home field, the priority of "closing a window" in the operation category with extremely short response is higher than the priority of "opening an air conditioner" and "opening a television" in the operation category with extremely short response is higher than the priority of "playing a tv series with a certain name", and the operation categories without obvious sequence can be set as the same level, for example, people propose "closing a window and opening a television". "the machine end can randomly respond to the condition that the machine end only is suitable for one round and only can respond to one type? "or" what is also required to ask for a request after the television has been turned on? ". In the case that the machine side is suitable for the combined response, the combined response can be from high to low according to the priority, namely that the window is closed and the air conditioner is opened. "," the television has been turned on and a television play of a certain name starts. "," the window has been closed and the television has been opened. ".
For example, the field of hastening, comforting the emotional category, "why your incoming number was marked as a fraud" was a higher priority than "last staff served very badly," which was higher than "your very badly served". When the people end states that' last staff has poor service, your service is not good, and the incoming call numbers of your are marked as fraud calls. "the machine end can reply" sorry to you, some customers do not want to be owing and mark the number as a fraud call, and we just remind you to pay back to the account you consume in time when only one round of response can be carried out. "; under the condition that the machine end is suitable for combined response, the combined response from high to low can be 'sorry to bother you, some customers do not want to arrear and mark numbers as fraud calls', the daily work at the side of us is to remind the customers to pay back to accounts which are consumed in time, and overdue amount bank can collect interest according to convention. The last staff who feeds back and my service attitude question, i have registered the follow-up to feed back to the superior processing in time. ".
The above-mentioned semantic groups and the response mode of the matching template set corresponding to the semantic groups, the set monolingual group response or the multilingual group response are combined by setting the relative multi-layer priority of each preset semantic group in the preset semantic group set, so that the response of the semantic groups to the linguisticized natural language semantic comprehension of the human port and the controllable categories after comprehension in the actual service can be flexibly met.
Step S43: and combining the matching template set corresponding to each preset basic semantic group in the preset basic semantic group set with the two layers of priorities to form a preset basic template group set.
Fig. 7 is a method for semantic understanding processing of a short text disclosed in the embodiment of the present application, including the following steps:
step M1: and generating each preset item semantic group set and a corresponding preset item template group set. The detailed procedure of step M1 is described below in fig. 8.
Step M2: and acquiring the short text to be semantically understood.
Step M3: and preprocessing the short text to be semantically understood.
Step M4: and matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching. The number of the preset item semantic groups in the matching is consistent with the number of the preset item semantic groups contained in the short text to be semantically understood.
The preset project semantic group set is [ PS1, PS2, ·, PSn ], where PS1 represents the 1 st semantic group of the preset project, PS2 represents the 2 nd semantic group of the preset project, … …; n is the number of preset item semantic groups in the preset item semantic group set;
a preset item template group set corresponding to the preset item semantic group set is [ PT1, PT 2., PTn ];
there is a one-to-one correspondence between PSi, i 1, 2.
When the matching method is used, the single item semantic group in the matching refers to a preset item template group corresponding to the single preset item semantic group in the matching, for example, a preset item semantic group PSi (i is 1, 2., n) in the matching refers to a preset item template group PTi corresponding to the preset item semantic group PSi in the matching, and specifically refers to the following three cases.
More than or equal to one rule template in a formal class template set in a preset item template set PTi corresponding to the preset item semantic group PSi in the matching;
more than or equal to one rule template in an auxiliary class template set in a preset item template set PTi corresponding to the preset item semantic group PSi in the matching;
and more than or equal to one rule template in a formal class template set and more than or equal to one rule template in an auxiliary class template set in a preset item template set PTi corresponding to the preset item semantic group PSi in the matching.
When the matching method is used, the multiple item semantic groups in the matching refer to the preset item template groups corresponding to the multiple preset item semantic groups in the matching, for example, the preset item semantic groups PSj, PSk, PSm, j ≠ k ≠ m in the matching refers to the preset item template groups PTj, PTk, and PTm corresponding to the preset item semantic groups PSj, PSk, and PSm in the matching, and specifically refers to more than or equal to one rule template in the respective formal template sets PTj, PTk, and PTm corresponding to the preset item semantic groups PSj, PSk, and PSm in the matching, and more than or equal to zero rule templates in the auxiliary template set;
step M5: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode. The corresponding project response modes refer to the following two types:
responding according to a single preset item semantic group. And directly responding to the single matched preset item semantic group according to the preset response content of the preset item semantic group. And when a plurality of preset item semantic groups are matched, selecting the semantic group with the highest priority according to the priority of the preset item semantic groups, and responding according to the corresponding response content.
Responding according to the number of the preset item semantic groups in the actual matching. And the single semantic group in the matching directly responds according to the preset response content of the semantic group. And when the medium and multi-semantic groups are matched, the responses are combined in sequence from high priority to low according to the priority of each semantic group and the corresponding response content.
Fig. 8 is a flowchart for generating a project template set in the short text semantic understanding processing method disclosed in the embodiment of the present application, including the following steps:
step M11: and classifying the minimum semantic unit in the semantic library into a plurality of preset item semantic groups with preset semantic widths according to semantic classification corresponding to the preset response range of each item, and generating a preset item semantic group set corresponding to each item.
For example, the preset response range of the item P3 is 3 replies (e.g., a31, a32, a33), each reply having a semantic classification including a plurality of minimum semantic units (e.g.,) There are 3 preset item semantic groups (e.g., PS31, PS32, PS33) for the item, and the minimum semantic unit included in each semantic group is consistent with the minimum semantic unit in the semantic classification corresponding to each sentence reply in the preset item response range (e.g.,)。
step M12: and generating one-to-one or one-to-many corresponding relation between each preset item semantic group and the preset basic semantic group in the preset item semantic group set of the corresponding item by combining the preset basic semantic group set according to the preset item semantic group set of each item.
For example, PS11 in the item semantic group set of the item P1 includes the minimum semantic unit S1+ S2+ S3, BS1 in the base semantic group set includes the minimum semantic unit S1, BS2 in the base semantic group set includes the minimum semantic unit S2+ S3, and then the corresponding relationship between PS11 and the base semantic group in the base semantic group set is one-to-two, specifically, one-to-two
The preset item semantic group can correspond to more than or equal to a preset basic semantic group; if a certain project requires a subdivided preset project semantic group, generating a correspondingly subdivided preset basic semantic group, and ensuring that the preset project semantic group can correspond to more than or equal to one preset basic semantic group; and different projects can generate a customized preset project semantic group set corresponding to the project and a corresponding relation between each preset project semantic group and a preset basic semantic group in the preset project semantic group set customized corresponding to the project.
Step M13: and presetting a preset basic module group set corresponding to the basic semantic group set by combining the priority and the corresponding relation between each preset item semantic group in each preset item semantic group set and each preset basic semantic group in the preset basic semantic group set, and generating a preset item module group set corresponding to each preset item semantic group set.
When establishing the 2-layer priority of the preset item semantic group set according to the corresponding relation between each preset item semantic group in the preset item semantic group set and the preset basic semantic group, if the corresponding relation is one-to-one, the priority of the preset item semantic group is consistent with the priority of the corresponding preset basic semantic group, and if one preset item semantic group corresponds to a plurality of preset basic semantic groups, the 2 layer with the lowest priority in the corresponding preset basic semantic groups is selected as the priority of the preset item semantic group.
The corresponding relation between the preset item semantic group set and the preset basic semantic group set is consistent with the corresponding relation between the preset item template group set and the preset basic template group set.
And different projects can generate a preset project template group set corresponding to the customized preset project semantic group set corresponding to the project.
By setting the preset basic semantic group and the preset project semantic group, according to the many-to-one correspondence between the preset basic semantic group and the preset project semantic group, the template group set corresponding to one set of basic semantic group set can be used for a plurality of projects, and can make corresponding response according to the response mode of project requirements, so that the actual service requirements can be met, and the production efficiency is improved.
In a corresponding short text semantic understanding processing link in a human-end and machine-end interactive scene, performing short text semantic understanding processing by using a corresponding preset item semantic group set and a corresponding preset item template group set, and collecting short texts to be semantically understood which are not repeated with the collected short texts. And (3) optimizing a semantic library by using the collected short text which is not repeated with the collected business requirement and is to be semantically understood, optimizing the generated preset item semantic group set, further optimizing a preset basic semantic group set, a preset basic template group set and each preset item template group set.
The semantic understanding template generation method and the semantic understanding processing method in the embodiment support semantic understanding processing when a small amount of data exists and multiple response modes of a machine end during man-machine interaction, and can continuously improve the accuracy of natural language semantic understanding processing in interaction along with the increase of data volume.
Fig. 9 is a short text semantic understanding template generating apparatus disclosed in the embodiment of the present application, which includes a short text collecting unit, a basic semantic group set generating unit, a template generating unit, and a basic template group set generating unit, where functions of the units are described as follows:
a short text collection unit: collecting a first preset number of short texts to be semantically understood;
a basic semantic group generating unit: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set consists of a plurality of preset basic semantic groups which do not have inclusion or included relationship;
a template generation unit: generating and/or optimizing a rule template set based on a first preset number of short texts to be semantically understood, and further generating and/or optimizing a matching template set corresponding to a preset basic semantic group set, wherein the rule template comprises a plurality of regular template components and is a minimum unit for performing short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the matching templates are divided into a formal class and an auxiliary class, and each class consists of a plurality of corresponding rule templates;
a basic template group set generation unit: and the matching template set is combined with the priority to generate a preset basic template set corresponding to the preset basic semantic set.
Fig. 10 is a short text semantic understanding processing apparatus disclosed in the embodiment of the present application, including an item template set generation unit, a short text acquisition unit, a short text preprocessing unit, a matching unit, and a response unit, where functions of the units are described as follows:
a project template group generating unit: generating each preset item semantic group set and a corresponding preset item template group set;
a short text acquisition unit: acquiring a short text to be semantically understood;
a short text preprocessing unit: preprocessing short texts to be semantically understood;
a matching unit: matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching;
a response unit: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode.
The embodiments of the apparatuses described in fig. 9 and 10 above are merely schematic, where the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual requirements to achieve the purpose of the solution of the embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Fig. 11 is a structure of a device disclosed in an embodiment of the present application, including a processor, a memory, a communication bus, and a short text semantic understanding processing interface for each item. The function of each part is explained as follows:
a processor: the method comprises the steps of receiving and executing a computer program in a memory, and realizing semantic understanding template generation and semantic understanding processing;
a memory: a computer program is stored which is executable on a processor.
In addition, the content stored in the memory may be implemented in the form of a software functional unit and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Therefore, the technical solution of the present application may substantially contribute to the prior art or may be embodied in the form of a software product stored in a storage medium, and including several documents and several logic instructions for enabling a device (which may be a computer, a server, a mobile phone, or a network device) with an operating system to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing files and program codes.
Communication bus: and the processor, the memory and the project short text semantic understanding processing interface are responsible for communication among each other.
Each item short text semantic understanding processing interface: and outputting the semantic understanding result to the corresponding module of the intelligent interaction equipment.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (14)
1. A method for generating a short text semantic understanding template is characterized by comprising the following steps:
step S1: collecting a first preset number of short texts to be semantically understood;
step S2: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set consists of a plurality of preset basic semantic groups which do not have inclusion or included relationship;
step S3: generating and/or optimizing a rule template set based on a first preset number of short texts to be semantically understood, and further generating and/or optimizing a matching template set corresponding to a preset basic semantic group set, wherein the rule template comprises a plurality of regular template components and is a minimum unit for performing short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the matching templates are divided into a formal class and an auxiliary class, and each class consists of a plurality of corresponding rule templates;
step S4: and the matching template set is combined with the priority to generate a preset basic template set corresponding to the preset basic semantic set.
2. The method according to claim 1, wherein the step S2 is configured to generate a preset basic semantic group set according to a first preset number of short texts to be semantically understood, and includes:
step S21: generating a semantic library, wherein the semantic library consists of all nonrepeating minimum semantic units in a first preset number of short texts to be semantically understood; the minimum semantic unit refers to the semantics of a minimum number of clause combinations; the clauses refer to Chinese punctuations connected with text contents and then Chinese punctuations; the minimum number of clause combinations means that the semantics of the clause combinations smaller than the minimum number in the natural language expression cannot be understood and therefore cannot be reasonably responded, and the semantics of the clause combinations reaching the corresponding minimum number can be understood and responded;
step S22: and classifying the minimum semantic unit in the semantic library into a plurality of preset basic semantic groups with preset semantic widths according to semantic classification corresponding to a response range preset by the service or the field, and generating a preset basic semantic group set.
3. The method according to claim 1, wherein the step S3 is to generate and/or optimize a rule template set based on a first preset number of short texts to be semantically understood, and further generate and/or optimize a matching template set corresponding to a preset base semantic group set, and includes:
step S31: selecting a third preset number of short texts to be semantically understood from the first preset number of short texts to be semantically understood, wherein the third preset number of short texts to be semantically understood simultaneously meets 2 conditions of related minimum semantic units S1, and the minimum semantic units S1 are classified in a preset basic semantic group BSi;
the 2 conditions are:
the condition is that the short text to be semantically understood contains the semantics of the minimum semantic unit S1;
the second condition is that the semantic of the minimum semantic unit S1 is consistent with the overall semantic of the whole short text to be semantically understood;
step S32: extracting key characters in a plurality of minimum length combinations for determining the semantics of a minimum semantic unit S1 from a third preset number of short texts to be semantically understood, wherein the length refers to the number of the key characters; the minimum length combination is divided into a formal class and an auxiliary class according to whether ambiguity is generated during understanding;
step S33: selecting a proper regular template component according to the types of the key characters in each extracted minimum length combination, and generating and/or optimizing a rule template set corresponding to a minimum semantic unit S1 by combining the relative positions of the key characters, wherein the types of the key characters comprise synonym characters, forbidden characters, synonym out-of-order n-time common characters, and 4 types of special characters which are partially or completely forbidden;
step S34: according to the type of each minimum length combination, putting the corresponding rule template into the corresponding formal class or auxiliary class set to form a set of preset matching templates corresponding to the minimum semantic unit S1;
step S35: sequentially repeating the steps for other minimum semantic units contained in the preset basic semantic group BSi to generate a preset matching template set of the preset basic semantic group BSi;
step S36: and circulating the steps to generate a preset matching template set of all preset basic semantic groups in the preset basic semantic group set.
4. The method according to claim 3, wherein the step S33 selects a suitable regular template component according to the extracted category of the key characters in each minimum length combination, and generates and/or optimizes a rule template set corresponding to the minimum semantic unit S1 according to the relative positions of the key characters, including:
step S331: preprocessing one short text to be semantically understood in a third preset number of short texts to be semantically understood;
step S332: finding out a minimum number of clause combinations which accord with the semantics of the whole short text and contain key characters in the found minimum length combinations from the preprocessed short text to be semantically understood;
step S333: selecting a corresponding regular template component according to the key character type in each minimum length combination in the minimum number of clause combinations;
step S334: combining the relative positions of the key characters in the minimum number of clause combinations found out and the use positions of the selected regular template components to generate or optimize a preset rule template;
step S335: repeating the above processes for a third preset number of short texts to be semantically understood, generating and/or optimizing a plurality of rule templates corresponding to the minimum semantic unit S1 in each short text, and combining the rule templates into a rule template set corresponding to the minimum semantic unit S1.
5. The method according to claim 4, wherein the step S334, in combination with the relative positions of the key characters in the found minimum number of clause combinations, the usage positions of the selected regular template components, generates or optimizes the preset rule template, including:
searching whether a preset rule template exists in a formal template set and an auxiliary template set in a preset basic matching template corresponding to a corresponding preset basic semantic group to meet the requirement of a selected regular template component and a position connection structure thereof, wherein the position connection structure refers to the relative position of each part of the rule template determined by the relative position of a key character and the use position of the regular template component;
if the applicable preset rule template exists, the key characters are filled in the corresponding positions, and the applicable preset rule template is optimized, so that the preset rule template can be matched with the preprocessed short text to be semantically understood;
and if no applicable preset rule template exists, filling the key characters into the corresponding regular template components, and connecting and generating the applicable preset rule template according to the relative positions of the key characters and the using positions of the components, so that the preset rule template can be matched with the preprocessed short text to be semantically understood.
6. The method according to claim 3, wherein the step S34 is to put the corresponding rule template into the corresponding formal class or auxiliary class set according to the category of each minimum length combination, and form a set of preset matching templates corresponding to the minimum semantic unit S1, and includes:
step S341: putting a rule template formed by key characters in the formal class minimum length combination into a corresponding formal class template set;
step S342: putting a rule template formed by key characters in the auxiliary class minimum length combination into a corresponding auxiliary class template set;
step S343: and combining the formal template set and the auxiliary template set to form a preset matching template set corresponding to the minimum semantic unit S1.
7. The method according to claim 1, wherein the step S4 of generating a preset base template set corresponding to a preset base semantic set by combining the matching template set with priorities includes:
step S41: dividing all preset basic semantic groups into a plurality of large categories according to response content, modes and purposes required by services or fields, determining the priority among the large categories, and generating a preset layer 1 priority;
step S42: determining the priority among preset basic semantic groups in each large category according to an actual reasonable response sequence, and generating a preset layer 2 priority;
step S43: and combining the matching template set corresponding to each preset basic semantic group in the preset basic semantic group set with the 2-layer priority to form a preset basic template group set.
8. A short text semantic understanding processing method is characterized by comprising the following steps:
step M1: generating each preset item semantic group set and a corresponding preset item template group set;
step M2: acquiring a short text to be semantically understood;
step M3: preprocessing short texts to be semantically understood;
step M4: matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching;
step M5: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode.
9. The method according to claim 8, wherein the step M1 of generating each preset item semantic group set and the corresponding preset item template group set comprises:
step M11: classifying the minimum semantic unit in the semantic library into a plurality of preset item semantic groups with preset semantic widths according to semantic classification corresponding to the preset response range of each item, and generating a preset item semantic group set corresponding to each item;
step M12: generating one-to-one or one-to-many corresponding relation between each preset item semantic group and a preset basic semantic group in the preset item semantic group set of the corresponding item by combining the preset basic semantic group set according to the preset item semantic group set of each item;
step M13: and combining the priority, the corresponding relation between each preset item semantic group in each preset item semantic group set and each preset basic semantic group in the preset basic semantic group set, presetting the preset basic module group set corresponding to the basic semantic group set, and generating the preset item module group set corresponding to each preset item semantic group set.
10. The method according to any one of claims 1 to 9, further comprising:
in a corresponding short text semantic understanding processing link in a human-end and machine-end interactive scene, performing short text semantic understanding processing by using a corresponding preset item semantic group set and a corresponding preset item template group set, and collecting short texts to be semantically understood which are not repeated with the collected short texts; and optimizing the generated preset item semantic group set by using the collected and collected nonrepeated short texts to be semantically understood in combination with the continuously adjusted business requirements, and further optimizing the preset basic semantic group set, the preset basic template group set and each preset item template group set.
11. A semantic understanding template generation apparatus, comprising:
a short text collection unit: collecting a first preset number of short texts to be semantically understood;
a basic semantic group generating unit: generating a preset basic semantic group set according to a first preset number of short texts to be semantically understood, wherein the preset basic semantic group set consists of a plurality of preset basic semantic groups which do not have inclusion or included relationship;
a template generation unit: generating and/or optimizing a rule template set based on a first preset number of short texts to be semantically understood, and further generating and/or optimizing a matching template set corresponding to a preset basic semantic group set, wherein the rule template comprises a plurality of regular template components and is a minimum unit for performing short text semantic understanding matching processing; the regular template component comprises a regular expression, punctuation combination and key characters, and is limited to be used at a specified position of the regular template; the matching templates are divided into a formal class and an auxiliary class, and each class consists of a plurality of corresponding rule templates;
a basic template group set generation unit: and the matching template set is combined with the priority to generate a preset basic template set corresponding to the preset basic semantic set.
12. A semantic understanding processing apparatus, comprising:
a project template group generating unit: generating each preset item semantic group set and a corresponding preset item template group set;
a short text acquisition unit: acquiring a short text to be semantically understood;
a short text preprocessing unit: preprocessing short texts to be semantically understood;
a matching unit: matching and querying the preprocessed short text to be semantically understood and a preset item template set corresponding to the preset item semantic set to obtain a preset item semantic set subset in matching;
a response unit: and according to the preset item semantic group subset in the matching, making a corresponding response to the human end by combining a corresponding item response mode.
13. An electronic device comprising a processor, a memory, and a computer program stored on and executable on the memory, a communication bus, and a short text semantic understanding processing interface for each item, wherein the processor implements the method of any one of claims 1 to 10 when executing the program.
14. A computer readable storage medium having stored thereon executable instructions, wherein the executable instructions when executed by a processor implement the method of any one of claims 1 to 10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011359958.1A CN112395885B (en) | 2020-11-27 | 2020-11-27 | Short text semantic understanding template generation method, semantic understanding processing method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011359958.1A CN112395885B (en) | 2020-11-27 | 2020-11-27 | Short text semantic understanding template generation method, semantic understanding processing method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112395885A true CN112395885A (en) | 2021-02-23 |
CN112395885B CN112395885B (en) | 2024-01-26 |
Family
ID=74604674
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011359958.1A Active CN112395885B (en) | 2020-11-27 | 2020-11-27 | Short text semantic understanding template generation method, semantic understanding processing method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112395885B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115563964A (en) * | 2022-11-10 | 2023-01-03 | 北京泰迪熊移动科技有限公司 | Regular short message text generation method and device, electronic equipment and storage medium |
CN116090789A (en) * | 2023-03-03 | 2023-05-09 | 麦高(广东)数字科技有限公司 | Lean manufacturing production management system and method based on data analysis |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107562732A (en) * | 2017-10-26 | 2018-01-09 | 北京康夫子科技有限公司 | The processing method and system of electronic health record |
CN108287858A (en) * | 2017-03-02 | 2018-07-17 | 腾讯科技(深圳)有限公司 | The semantic extracting method and device of natural language |
WO2018153295A1 (en) * | 2017-02-27 | 2018-08-30 | 腾讯科技(深圳)有限公司 | Text entity extraction method, device, apparatus, and storage media |
CN109284363A (en) * | 2018-12-03 | 2019-01-29 | 北京羽扇智信息科技有限公司 | A kind of answering method, device, electronic equipment and storage medium |
CN110069623A (en) * | 2017-12-06 | 2019-07-30 | 腾讯科技(深圳)有限公司 | Summary texts generation method, device, storage medium and computer equipment |
RU2697647C1 (en) * | 2018-10-01 | 2019-08-15 | Общество с ограниченной ответственностью "Аби Продакшн" | System and method for automatic creation of templates |
EA201800581A1 (en) * | 2018-09-24 | 2020-03-31 | Общество С Ограниченной Ответственностью "Незабудка Софтвер" | SEARCH METHOD IN TEXT OF MATCHES WITH PATTERNS |
CN111144128A (en) * | 2019-12-26 | 2020-05-12 | 北京百度网讯科技有限公司 | Semantic parsing method and device |
CN111292751A (en) * | 2018-11-21 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Semantic analysis method and device, voice interaction method and device, and electronic equipment |
CN111460123A (en) * | 2020-04-07 | 2020-07-28 | 中国搜索信息科技股份有限公司 | Conversation intention identification method and device for teenager chat robot |
CN111581329A (en) * | 2020-04-23 | 2020-08-25 | 上海兑观信息科技技术有限公司 | Short text matching method and device based on inverted index |
CN111858900A (en) * | 2020-09-21 | 2020-10-30 | 杭州摸象大数据科技有限公司 | Method, device, equipment and storage medium for generating question semantic parsing rule template |
CN112395394A (en) * | 2020-11-27 | 2021-02-23 | 安徽迪科数金科技有限公司 | Short text semantic understanding template inspection method, template generation method and device |
-
2020
- 2020-11-27 CN CN202011359958.1A patent/CN112395885B/en active Active
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018153295A1 (en) * | 2017-02-27 | 2018-08-30 | 腾讯科技(深圳)有限公司 | Text entity extraction method, device, apparatus, and storage media |
CN108287858A (en) * | 2017-03-02 | 2018-07-17 | 腾讯科技(深圳)有限公司 | The semantic extracting method and device of natural language |
WO2018157703A1 (en) * | 2017-03-02 | 2018-09-07 | 腾讯科技(深圳)有限公司 | Natural language semantic extraction method and device, and computer storage medium |
CN107562732A (en) * | 2017-10-26 | 2018-01-09 | 北京康夫子科技有限公司 | The processing method and system of electronic health record |
CN110069623A (en) * | 2017-12-06 | 2019-07-30 | 腾讯科技(深圳)有限公司 | Summary texts generation method, device, storage medium and computer equipment |
EA201800581A1 (en) * | 2018-09-24 | 2020-03-31 | Общество С Ограниченной Ответственностью "Незабудка Софтвер" | SEARCH METHOD IN TEXT OF MATCHES WITH PATTERNS |
RU2697647C1 (en) * | 2018-10-01 | 2019-08-15 | Общество с ограниченной ответственностью "Аби Продакшн" | System and method for automatic creation of templates |
CN111292751A (en) * | 2018-11-21 | 2020-06-16 | 北京嘀嘀无限科技发展有限公司 | Semantic analysis method and device, voice interaction method and device, and electronic equipment |
CN109284363A (en) * | 2018-12-03 | 2019-01-29 | 北京羽扇智信息科技有限公司 | A kind of answering method, device, electronic equipment and storage medium |
CN111144128A (en) * | 2019-12-26 | 2020-05-12 | 北京百度网讯科技有限公司 | Semantic parsing method and device |
CN111460123A (en) * | 2020-04-07 | 2020-07-28 | 中国搜索信息科技股份有限公司 | Conversation intention identification method and device for teenager chat robot |
CN111581329A (en) * | 2020-04-23 | 2020-08-25 | 上海兑观信息科技技术有限公司 | Short text matching method and device based on inverted index |
CN111858900A (en) * | 2020-09-21 | 2020-10-30 | 杭州摸象大数据科技有限公司 | Method, device, equipment and storage medium for generating question semantic parsing rule template |
CN112395394A (en) * | 2020-11-27 | 2021-02-23 | 安徽迪科数金科技有限公司 | Short text semantic understanding template inspection method, template generation method and device |
Non-Patent Citations (4)
Title |
---|
KENT BIMSON ET AL: "Find-to-Forecast Process: An Automated Methodology for Situation Assessment", 《2014 IEEE INTERNATIONAL CONFERENCE ON SEMANTIC COMPUTING》, pages 1 - 5 * |
张亚飞 等: "基于词性软模板注意力机制的短文本自动摘要方法", 《模式识别与人工智能》, vol. 33, no. 6, pages 551 - 558 * |
曹雨 等: "中文复述问句生成技术研究", 《智能计算机与应用》, vol. 9, no. 1, pages 192 - 198 * |
符磊: "基于文本分类的信息解析研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, no. 7, pages 138 - 1473 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115563964A (en) * | 2022-11-10 | 2023-01-03 | 北京泰迪熊移动科技有限公司 | Regular short message text generation method and device, electronic equipment and storage medium |
CN116090789A (en) * | 2023-03-03 | 2023-05-09 | 麦高(广东)数字科技有限公司 | Lean manufacturing production management system and method based on data analysis |
CN116090789B (en) * | 2023-03-03 | 2023-08-29 | 麦高(广东)数字科技有限公司 | Lean manufacturing production management system and method based on data analysis |
Also Published As
Publication number | Publication date |
---|---|
CN112395885B (en) | 2024-01-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20210357768A1 (en) | Electronic Meeting Intelligence | |
Xu et al. | Long time no see! open-domain conversation with long-term persona memory | |
US10679134B2 (en) | Automated ontology development | |
US10268990B2 (en) | Electronic meeting intelligence | |
CN113704451B (en) | Power user appeal screening method and system, electronic device and storage medium | |
Shrestha et al. | Detection of question-answer pairs in email conversations | |
CN110825858A (en) | Intelligent interaction robot system applied to customer service center | |
CN110462730A (en) | Promote with the end-to-end communication of multilingual and automation assistant | |
CN108780527A (en) | For determining reply content based on the schedule data of the feature of electronic communication and user to be included in user to the method and apparatus in the reply of electronic communication | |
US10860566B1 (en) | Themes surfacing for communication data analysis | |
CN112434501B (en) | Method, device, electronic equipment and medium for intelligent generation of worksheet | |
US11170184B2 (en) | Computer implemented system and method for automatically generating messages | |
CN110493019A (en) | Automatic generation method, device, equipment and the storage medium of meeting summary | |
CN112395885A (en) | Short text semantic understanding template generation method, semantic understanding processing method and device | |
CN110222145A (en) | A kind of intelligence law appraisal procedure and system | |
US11966698B2 (en) | System and method for automatically tagging customer messages using artificial intelligence models | |
CN107341157B (en) | Customer service conversation clustering method and device | |
US20240005089A1 (en) | Document auto-completion | |
CN112287082A (en) | Data processing method, device, equipment and storage medium combining RPA and AI | |
CN112395394B (en) | Short text semantic understanding template checking method, template generating method and device | |
CN109919657A (en) | User demand information acquisition method and device, storage medium and voice equipment | |
CN110633123A (en) | Intelligent calendar system based on mobile terminal or PC terminal | |
CN116578682B (en) | Intelligent consultation method and system for legal service | |
CN111428018B (en) | Intelligent question-answering method and device | |
US20220138426A1 (en) | Electronic device, method, and computer program which support naming |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |