CN111078855A - Information processing method, information processing device, electronic equipment and storage medium - Google Patents
Information processing method, information processing device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111078855A CN111078855A CN201911316145.1A CN201911316145A CN111078855A CN 111078855 A CN111078855 A CN 111078855A CN 201911316145 A CN201911316145 A CN 201911316145A CN 111078855 A CN111078855 A CN 111078855A
- Authority
- CN
- China
- Prior art keywords
- intention
- label
- tag
- target text
- elements
- 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.)
- Pending
Links
Images
Classifications
-
- 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
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- 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
- G06F16/355—Class or cluster creation or modification
-
- 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
- G06F16/38—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Finance (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Library & Information Science (AREA)
- Machine Translation (AREA)
Abstract
The application discloses an information processing method, an information processing device, electronic equipment and a storage medium, wherein tag elements are obtained by splitting obtained target texts; then, performing intention label matching on the label element through a pre-created intention database to obtain a first intention label; an intent tag that matches the target text is determined based on the intent tag. Compared with the prior art that the intention of the target text is directly identified, the splitting process of the label element is added in the intention identification process, the purpose that the label element is used for matching the intention label is achieved, the judgment basis for analyzing the intention of the input target text is used, and the problems of limitation of complicated text label intention classification and inaccurate intention analysis can be solved.
Description
Technical Field
The present application relates to the field of information processing technologies, and in particular, to an information processing method and apparatus, and an electronic device.
Background
The intelligent customer service system is an automatic service system for industrial application developed on the basis of large-scale knowledge processing, and establishes an intelligent communication path based on natural language processing for communication between a user and a service provider.
In intelligent customer service systems, natural language understanding techniques are typically employed to map user questions onto defined intents based on the user's input text for a single turn or for multiple turns. However, the existing intelligent system has certain limitation when the input text is a complex label. For example, "i updated my mobile phone system" or "i could not update mobile phone system" often misclassify in the existing intelligent customer service system by using the classification model, which finally results in inaccurate analysis of the obtained user intention.
Disclosure of Invention
In view of this, the present application provides the following technical solutions:
an information processing method comprising:
acquiring a target text;
splitting the target text to obtain tag elements, wherein the tag elements represent key words with specific semantics;
querying in an intention database by using the label element to obtain a first intention label matched with the label element, wherein the intention database comprises a label element set and an intention label set, and the label element set and the intention label set have a preset mapping relation;
based on the first intent tag, an intent tag that matches the target text is determined.
Optionally, the method further comprises:
inputting the label elements into a pre-constructed first identification model, and determining predicted intention labels corresponding to the label elements through the first identification model, wherein the first identification model has the capacity of enabling the intention labels corresponding to the label elements to trend to actual intention labels corresponding to the label elements;
and creating an intention database according to the label elements and the intention labels matched with the label elements.
Optionally, the method further comprises:
obtaining sample information, the sample information including a label element and an intention label;
and respectively taking the obtained sample information as the training input of the neural network, and training to obtain a first recognition model.
Optionally, the determining an intention tag matching the target text based on the first intention tag includes:
performing intention identification on the target text to obtain a second intention label;
and performing fusion processing on the first intention label and the second intention label to obtain an intention label of the target text.
Optionally, the performing intent recognition on the target text to obtain a second intent tag includes:
inputting the target text into a pre-constructed second recognition model, and determining a predicted second intention label corresponding to the target text through the second recognition model;
wherein the second recognition model has the capability of trending a second intention label corresponding to the target text towards an actual intention label corresponding to the target text; and the second recognition model is a model obtained by training by taking the obtained sample information as the training input of the neural network, wherein the sample information is the information matched with the target text information.
Optionally, the splitting the tag element of the target text to obtain the tag element includes:
and extracting the features of the target text, and determining the label elements according to the extracted features.
Optionally, the splitting the tag element of the target text to obtain the tag element includes:
classifying the target text by utilizing a pre-established classification model to obtain a label element; and the pre-created classification model representation judges whether the target text contains a model of a specific tag element.
An information processing apparatus comprising:
an acquisition unit configured to acquire a target text;
the splitting unit is used for splitting the label elements of the target text to obtain the label elements, and the label elements represent key words with specific semantics;
the query unit is used for querying in an intention database by using the tag elements to obtain a first intention tag matched with the tag elements, the intention database comprises a tag element set and an intention tag set, and the tag element set and the intention tag set have a preset mapping relation;
a determining unit, configured to determine, based on the first intention tag, an intention tag that matches the target text.
An electronic device, comprising:
a memory for storing a program;
a processor configured to execute the program, the program specifically configured to:
acquiring a target text;
splitting the target text to obtain tag elements, wherein the tag elements represent key words with specific semantics;
querying in an intention database by using the label element to obtain a first intention label matched with the label element, wherein the intention database comprises a label element set and an intention label set, and the label element set and the intention label set have a preset mapping relation;
based on the first intent tag, an intent tag that matches the target text is determined.
A storage medium, characterized in that the storage medium stores computer program code, which when executed implements an information processing method as described in any one of the above.
According to the technical scheme, the application discloses an information processing method, an information processing device, electronic equipment and a storage medium, wherein the label element is obtained by splitting the obtained target text; then, performing intention label matching on the label element through a pre-created intention database to obtain a first intention label; an intent tag that matches the target text is determined based on the intent tag. Compared with the prior art that the intention of the target text is directly identified, the splitting process of the label element is added in the intention identification process, the purpose that the label element is used for matching the intention label is achieved, the judgment basis for analyzing the intention of the input target text is used, and the problems of limitation of complicated text label intention classification and inaccurate intention analysis can be solved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on the provided drawings without creative efforts.
FIG. 1 is a schematic diagram illustrating a scenario of an information processing system according to an embodiment of the present application;
FIG. 2 is a flow chart illustrating an information processing method according to an embodiment of the present disclosure;
FIG. 3 is a diagram illustrating a scenario for constructing an intention database according to an embodiment of the present application;
FIG. 4 is a flowchart illustrating a method for determining an intention label of a target text according to an embodiment of the present application;
FIG. 5 is a schematic diagram illustrating a target text splitting structure provided in an embodiment of the present application;
fig. 6 shows a schematic structural diagram of an information processing apparatus according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and 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.
The information processing method loaded in each embodiment of the present application may be applied to a scenario in which an intention analysis is performed on a text to be analyzed, for example, in a control architecture of an application system or an entity device corresponding to an intelligent customer service, a voice assistant, a question and answer system, an intelligent robot, or the like.
The information processing system provided by the embodiment of the application can be used for
Referring to fig. 1, a schematic diagram of a scenario of an information processing system according to an embodiment of the present application is shown. The information processing system 100 includes a server 110, a network 120, and a terminal 130.
The terminal 130 has information input and output functions, and can acquire and transmit information based on different information instructions. For example, the terminal 130 represents a mobile terminal, which may receive a target text through an information receiving module or some information processing application, and then transmit the target text to the server 110 through an information output module, so that the server 110 recognizes the target text, transmits a processed intention tag to the terminal 130, is output by the terminal 130, or outputs information matched with the intention tag. In some embodiments, the terminal 130 may include, but is not limited to, a smart phone, a tablet, a laptop, a smart robot, etc., or any combination thereof.
The following describes in detail an information processing method executed in the information processing system of the embodiment of the present application. The information processing method can be applied to the server 110 in the system, and can also be applied to a processor of the terminal 130.
Referring to fig. 2, a schematic flow chart of an information processing method provided in an embodiment of the present application is shown. The information processing method may include the steps of:
s201, acquiring a target text.
The target text characterizes the text that needs to be intent recognized. The original information format can be text, voice information or other information with any format. For example, when a user performs information interaction with a voice assistant function of the intelligent terminal, the user outputs voice information, and the intelligent terminal converts the voice information into text information after receiving the voice information, so that a processor of the intelligent terminal acquires the text information as a target text of subsequent processing. In some embodiments, the target text may represent a sentence composed of words, and the sentence may be a semantically complete sentence or a sentence with incomplete information, for example, a part of information input by the user is received, and may also be used as the target text in the embodiment of the present application.
In some embodiments, while the target text is obtained, text information having a semantic similar to that of the target text may also be obtained, for example, the target text input by the user is "how to update the mobile phone system", and correspondingly, while the target text is obtained, the target text may also be recorded as "how to update the mobile phone system". Therefore, the concrete representation format of the target text is not limited in the embodiment of the present application.
S202, carrying out label element splitting on the target text to obtain label elements.
The tag elements characterize keywords having a particular semantic meaning. I.e. the tag elements are part of the information of the target text, if the target text is a sentence, the tag elements may be words constituting the sentence. It should be noted that, the keywords with specific semantics represented by the tag elements still take the target text as an example, if the target text is "how to update my mobile phone," the corresponding tag element information may be "how," "update," or "mobile phone," and the "my" in the sentence is usually not taken as a tag element, because the semantic analysis or intent recognition of the entire target text is not completely affected by the word. That is, in the embodiment of the present application, the specific semantics are matched with the semantic information to be expressed by the target text.
When the label elements of the target text are split, the keywords of the target text can be extracted through a word segmentation algorithm; the high-frequency words can also be counted by a feature extraction method, and then the extracted feature words are matched with the high-frequency words to determine keywords, so that label elements are obtained; the target text can also be analyzed in a machine learning manner to obtain the tag elements corresponding to the target text. For a specific implementation of performing a tag element on a target text, reference may be made to descriptions in subsequent embodiments of the present application, which are not described herein again.
S203, inquiring in the intention database by using the label element to obtain a first intention label matched with the label element.
And S204, determining the intention label matched with the target text based on the first intention label.
In the embodiment of the application, an intention database is created in advance, the intention database comprises a tag element set and an intention tag set, and the tag element set and the intention tag set have a preset mapping relation. That is, there will be several intent tags in the intent database, and each intent tag will correspond to several tag elements. For example, the intention label is "how to update the system", and the corresponding label element set may be "how, update, system", so that three labels of "how", "update", and "system" are included in the label element set. After the tag elements of the target text are obtained, the tag elements can be used as query indexes to query in the intention database to obtain the intention tags.
In some embodiments, if a first intent tag is matched through the intent database and the first intent tag is the only intent tag corresponding to the tag element, the first intent tag may be the intent tag of the target text. In addition, the first intention label may be an intention label of the target text if the first intention label can reflect the intention of the target text.
In some embodiments, the tag elements may be part of a set of tag elements in the intent database. For example, still taking the above tag element set as an example, if the tag elements obtained by splitting the target text include "how" and "update", the intention tags corresponding to the above two tag elements may be matched, and at this time, the intention tags that may be matched are not unique, and further confirmation of the intention tags needs to be performed according to semantic information of the target text itself, or a context, and attribute information of the target text generation terminal or the user.
The embodiment of the application discloses an information processing method, which comprises the steps of splitting a tag element of an obtained target text to obtain the tag element; then, performing intention label matching on the label element through a pre-created intention database to obtain a first intention label; an intent tag that matches the target text is determined based on the intent tag. Compared with the prior art that the intention of the target text is directly identified, the splitting process of the label element is added in the intention identification process, the purpose that the label element is used for matching the intention label is achieved, the judgment basis for analyzing the intention of the input target text is used, and the problems of limitation of complicated text label intention classification and inaccurate intention analysis can be solved.
The manner in which various features of the embodiment shown in fig. 2 can be implemented is described below.
In another embodiment of the present application, a method for constructing an intention database is further provided, where the method forms an intention database by establishing a mapping relationship between a tag element and an intention tag, so that a corresponding intention tag can be conveniently found through the tag element in an application process, and the corresponding tag element can also be analyzed through the intention tag. The method for constructing intention data comprises the following steps:
s301, inputting the label element into a pre-constructed first identification model, and determining a predicted intention label corresponding to the label element through the first identification model.
S302, creating an intention database according to the label elements and the intention labels matched with the label elements.
Referring to fig. 3, a schematic diagram of a scenario in which an intention database is constructed according to an embodiment of the present application is shown. Firstly, a tag element 401 is obtained, where the tag element includes at least one tag element, that is, the tag element may be a tag element set, and the tag element in fig. 4 is composed of several tag elements, including: label element 1, label element 2 … … label element n, wherein each label element can represent one or more label elements, and each label in each label element has an association relationship, i.e. it can be derived from the same text information, and can also characterize the matching relationship in semantics. For example, tag element 1 includes tag element 10, tag element 11, and the like.
The obtained tag elements 401 are then input into the first recognition model 402, and the predicted intention tags corresponding to the tag elements are determined by the first recognition model 402. This first recognition model 402 has the ability to trend the intent tag to which the tag element corresponds to the actual intent tag to which the tag element corresponds.
The input information of the first recognition model 402 is a tag element 401, the output information thereof is an intention tag 403, and each intention tag 403 is matched with the input tag element 401. If the input is tag element 1, the output is intention tag 1. The tag elements 401 are then mapped one-to-one to the intent tags 403 to generate an intent database 404. The mapping relationship between the label elements and the intention labels is recorded in the intention database 404, so that the intention labels can be found in the intention database through the label elements, and the corresponding label elements can also be obtained through the intention labels.
For example, in FIG. 3 the intent database 404 may record information in the form of:
tag element 1{ how, update, system }, corresponding intent tag 1 "how to update system";
tag element 2{ unable, update, system }, corresponding intention tag 2 "unable to update system";
……
tag element n { how, update, apply }, corresponding intent tag 3 "how to update APP 1".
When constructing the intention database, the intention label is described at a higher level or at a lower level according to the application scenario of the user. If the obtained tag element includes { how, update, application }, if the tag element is obtained through the APP1 of the user currently applying, the application may be determined to be the APP1, that is, the obtained intention tag may be "how to update the APP 1". Correspondingly, a label including a specific application may be referred to as an application in the obtained intention label.
Correspondingly, in another embodiment of the present application, there is also provided a method of constructing a first recognition model, which may include the steps of:
and S501, obtaining sample information.
The sample information includes a label element and an intention label.
And S502, respectively taking the obtained sample information as training input of the neural network, and training to obtain a first recognition model.
In performing S501, the sample information may be information matching the target text. Firstly, a text corresponding to sample information is required to be obtained, wherein the text corresponding to the sample information is a text form such as a complete or partial sentence or paragraph formed by a plurality of words. The text corresponding to the sample information may be obtained in various ways. For example, by crawling text information from a webpage by using relevant technologies such as a crawler for texts historically input by a plurality of users, information format processing may be performed on a corresponding question and answer sentence in an existing intelligent interaction scene to obtain text information, and the like. The acquired text information is subsequently used for constructing sample information.
Then, a label is added to the obtained text information, in the embodiment of the present application, firstly, the text information needs to be subjected to label element extraction, and then, an intention label of the label element is determined. The label added to the text information represents whether the label element in the text information is matched with the intention label. For example, when a tag element in one text message does not fit to semantic information of an intention tag, a tag "1" may be added to the text message, and when a tag element in one text message fits to semantic information of an intention tag, a tag "0" may be added to the text message. The process enables the tagging of one or more constructed sample pairs, followed by a supervised training process by the user.
Next, when S502 is executed, the obtained sample information is used as training input of the neural network, and a first recognition model is obtained through training. The neural network used for training may include at least one of: deep Neural Networks (DNN) models, Convolutional Neural Networks (CNN) models, Recurrent Neural Networks (RNN) models, and the like, and may also include variations or migrations of the above various Neural Networks. The neural network may include one or more hidden layers. And S502 is executed, the neural network is trained by utilizing a plurality of sample pairs with labels, and parameters of the neural network are continuously optimized on the basis of the output of the neural network and the labels of the sample pairs until a target neural network is obtained, namely a first intention recognition model is obtained.
As can be understood by those skilled in the art, in the process of creating the first recognition model, a plurality of sample information is constructed, each sample information is constructed by a label element and an intention label, and the label is added according to the matching condition of the label element and the intention label in each sample; and then, carrying out supervised training on the neural network by utilizing a plurality of sample information with labels to obtain a first recognition model so as to obtain a recognition process of the intention labels according to the label elements, namely, the purpose of recognizing the intention according to part or all of the label elements in the text is realized based on deep learning.
In the above embodiment, the intention is identified through the tag elements included in the target text, and in order to obtain an intention tag more matched and more accurate with the target text, in the embodiment of the present application, the final intention tag of the target text may be determined by jointly analyzing according to the intention tag corresponding to the target text and the intention tag corresponding to the tag elements. Referring to fig. 4, which is a flowchart illustrating a method for determining a target text intention label according to an embodiment of the present application, the method may include the following steps:
s601, acquiring a target text;
s602, carrying out label element splitting on the target text to obtain label elements;
s603, inquiring in an intention database by using the label element to obtain a first intention label matched with the label element;
s604, performing intention identification on the target text to obtain a second intention label;
and S605, carrying out fusion processing on the first intention label and the second intention label to obtain the intention label of the target text.
In this embodiment, in the process of identifying the intention of the target text, a process of splitting a tag element is added in the existing intention identification process, that is, the target text may be split into a plurality of tag elements, and the tag elements are matched with the existing tag elements capable of corresponding to the first intention tag to obtain the first intention tag, where the first intention tag may assist in analyzing the intention of the target text. Then, the target text can be subjected to intention recognition by using the existing intention recognition mode to obtain a second intention label corresponding to the target text, and the first intention label and the second intention label are comprehensively analyzed to obtain the intention label of the target text. Therefore, even if the target text is incomplete, namely the intention of the target text cannot be obtained through single analysis, the intention label can be obtained through label meta-analysis corresponding to the target text.
When the first intention label and the second intention label are comprehensively analyzed, if the semantics of the two intention label representations are the same, any one of the intention labels can be selected as the final intention label of the target text. The weight values of the first intention label and the second intention label in the final result can also be determined according to the application environment where the target text input by the user is located or the intention recognition algorithm of the target text. For example, if the target text is incomplete, the final intention tag of the target text can be determined by the first intention tag.
In some embodiments, performing intent recognition on the target text to obtain a second intent tag includes: and inputting the target text into a pre-created second recognition model, and determining a predicted second intention label corresponding to the target text through the second recognition model.
Wherein the second intention recognition model has the capability of trending the second intention label corresponding to the target text towards the actual intention label corresponding to the target text. The second probabilistic model is a model obtained by training each piece of acquired sample information as a training input of the neural network. The sample information is information matched with the target text information.
The creation process of the second intention recognition model may refer to the creation process of the first recognition model, which is similar using a trained neural network. The process of adding labels to the constructed sample pairs is realized by whether the semantic information and the intention labels of the sample information are matched or not only when the sample information is created in the second intention identification model. The training process of the neural network is not described herein.
In another embodiment of the present application, a method of obtaining tag elements of a target text is also provided. In one possible implementation manner, feature extraction is performed on the target text, and the tag elements are determined according to the extracted features. The feature extracted by the feature extraction can represent information such as word features, semantic features and the like. For example, the word feature may be to determine whether the target text includes a target keyword, if so, extract a keyword in the target text that matches the target keyword, and determine the keyword as a tag element of the target text. That is, the tag element may be the target keyword itself or a synonym of the target keyword. For example, it may be determined whether the keyword "how" is included in the target text, and if so, the word is extracted as a tag element.
If the target text is subjected to feature extraction through semantic features, semantic information of the target text needs to be analyzed, whether the target text comprises words with preset semantics or not is judged, or word division can be performed according to the semantics, and divided words are obtained and serve as label elements. It should be noted that when the feature extraction is performed on the target text, the target text may be preprocessed first, for example, redundant subject information may be removed, and then the feature extraction is performed, which is beneficial to reducing the calculation complexity and the computation workload.
In another possible implementation manner, a pre-created classification model is used to classify the target text to obtain the tag elements, and the pre-created classification model represents and judges whether the target text contains a model of a specific tag element. The classification model can be used for providing classification basis for obtaining the label elements according to the target text. The classification model may comprise at least one classification system. The classification model may also include one or more classifiers obtained through training sample sets or machine learning. The machine learning algorithm may include, but is not limited to, at least one of the following: regression analysis methods, K-nearest neighbor methods, decision tree methods, neural network methods, naive bayes methods, support vector machine methods, and the like. The classification system in the classification model refers to classification conditions corresponding to different label elements.
The embodiments of the application can be applied to an intelligent question-answering system, if a target text input by a user is 'how to update a mobile phone system i wants to know', the target text is split to obtain a label element 'how to update, system', an intention label corresponding to the label element can be obtained by inquiring through an intention database and is 'how to update the system', and then the intelligent customer service can feed back relevant information about 'how to update the system' to the user according to the intention label.
For a corresponding process of analyzing the intention of the user target text, see an example table 701 corresponding to the target text splitting structure diagram shown in fig. 5, which is not described in this application.
In the embodiments of the present application, the problem of intent classification of a text facing a complex label can be solved, that is, firstly, a label element is split for a text with a complex label, that is, each text is split and decomposed into a plurality of label elements, then, an intent label is determined according to the label elements, since the label classification is at the bottom of intent recognition and the label elements and the intent label have a specified mapping relationship, a better bottom characteristic can be provided for the intent label classification by adding the process of label element classification, which is helpful for improving the classification result of an upper-layer intent label.
There is also provided in an embodiment of the present application an information processing apparatus, referring to fig. 6, the apparatus including:
an acquisition unit 801 configured to acquire a target text;
a splitting unit 802, configured to split a tag element of the target text to obtain a tag element, where the tag element represents a keyword with a specific semantic meaning;
the query unit 803 is configured to query, by using the tag element, in an intention database to obtain a first intention tag matched with the tag element, where the intention database includes a tag element set and an intention tag set, and the tag element set and the intention tag set have a preset mapping relationship;
a determining unit 804, configured to determine an intention label matching the target text based on the first intention label.
On the basis of the above embodiment, the apparatus further includes:
the database creating unit is used for inputting the label elements into a pre-constructed first identification model, and determining the predicted intention labels corresponding to the label elements through the first identification model, wherein the first identification model has the capacity of enabling the intention labels corresponding to the label elements to trend to the actual intention labels corresponding to the label elements;
and creating an intention database according to the label elements and the intention labels matched with the label elements.
On the basis of the above embodiment, the apparatus further includes:
a first model creating unit for obtaining sample information, the sample information including a label element and an intention label;
and respectively taking the obtained sample information as the training input of the neural network, and training to obtain a first recognition model.
On the basis of the above embodiment, the determining unit includes:
the text identification subunit is used for performing intention identification on the target text to obtain a second intention label;
and the fusion processing subunit is configured to perform fusion processing on the first intention tag and the second intention tag to obtain an intention tag of the target text.
On the basis of the foregoing embodiment, the text recognition subunit is specifically configured to:
inputting the target text into a pre-constructed second recognition model, and determining a predicted second intention label corresponding to the target text through the second recognition model;
wherein the second recognition model has the capability of trending a second intention label corresponding to the target text towards an actual intention label corresponding to the target text; and the second recognition model is a model obtained by training by taking the obtained sample information as the training input of the neural network, wherein the sample information is matched with the target text information.
On the basis of the above embodiment, the splitting unit includes:
and the feature extraction subunit is used for extracting features of the target text and determining the label element according to the extracted features.
On the basis of the above embodiment, the splitting unit further includes:
the classification subunit is used for classifying the target text by utilizing a pre-established classification model to obtain a label element; and the pre-created classification model representation judges whether the target text contains a model of a specific tag element.
The embodiment discloses an information processing device, which is used for splitting a tag element of an obtained target text by a splitting unit to obtain the tag element; then, performing intention label matching on the label elements in a query unit through a pre-created intention database to obtain a first intention label; an intention tag that matches the target text is determined in a determination unit based on the intention tag. Compared with the prior art that the intention of the target text is directly identified, the splitting process of the label element is added in the intention identification process, the purpose that the label element is used for matching the intention label is achieved, the judgment basis for analyzing the intention of the input target text is used, and the problems of limitation of complicated text label intention classification and inaccurate intention analysis can be solved.
There is also provided in an embodiment of the present application an electronic device, including:
a memory for storing a program;
a processor configured to execute the program, the program specifically configured to:
acquiring a target text;
splitting the target text to obtain tag elements, wherein the tag elements represent key words with specific semantics;
querying in an intention database by using the label element to obtain a first intention label matched with the label element, wherein the intention database comprises a label element set and an intention label set, and the label element set and the intention label set have a preset mapping relation;
based on the first intent tag, an intent tag that matches the target text is determined.
Further, the method further comprises:
inputting the label elements into a pre-constructed first identification model, and determining predicted intention labels corresponding to the label elements through the first identification model, wherein the first identification model has the capacity of enabling the intention labels corresponding to the label elements to trend to actual intention labels corresponding to the label elements;
and creating an intention database according to the label elements and the intention labels matched with the label elements.
Further, the method further comprises:
obtaining sample information, the sample information including a label element and an intention label;
and respectively taking the obtained sample information as the training input of the neural network, and training to obtain a first recognition model.
Further, the determining an intention tag matching the target text based on the first intention tag includes:
performing intention identification on the target text to obtain a second intention label;
and performing fusion processing on the first intention label and the second intention label to obtain an intention label of the target text.
Further, the performing intention recognition on the target text to obtain a second intention tag includes:
inputting the target text into a pre-constructed second recognition model, and determining a predicted second intention label corresponding to the target text through the second recognition model;
wherein the second recognition model has the capability of trending a second intention label corresponding to the target text towards an actual intention label corresponding to the target text; and the second recognition model is a model obtained by training by taking the obtained sample information as the training input of the neural network, wherein the sample information is the information matched with the target text information.
Further, the splitting the tag element of the target text to obtain the tag element includes:
and extracting the features of the target text, and determining the label elements according to the extracted features.
Further, the splitting the tag element of the target text to obtain the tag element includes:
classifying the target text by utilizing a pre-established classification model to obtain a label element; and the pre-created classification model representation judges whether the target text contains a model of a specific tag element.
A storage medium, characterized in that the storage medium stores computer program code, which when executed implements an information processing method as described in any one of the above.
The storage medium refers to a computer storage medium that may contain a propagated data signal with computer program code embodied therewith, for example, on a baseband or as part of a carrier wave. The propagated signal may take any of a variety of forms, including electromagnetic, optical, etc., or any suitable combination. A computer storage medium may be any computer-readable medium that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code located on a computer storage medium may be propagated over any suitable medium, including radio, cable, fiber optic cable, RF, or the like, or any combination of the preceding.
The emphasis of each embodiment in the present specification is on the difference from the other embodiments, and the same and similar parts among the various embodiments may be referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. An information processing method comprising:
acquiring a target text;
splitting the target text to obtain tag elements, wherein the tag elements represent key words with specific semantics;
querying in an intention database by using the label element to obtain a first intention label matched with the label element, wherein the intention database comprises a label element set and an intention label set, and the label element set and the intention label set have a preset mapping relation;
based on the first intent tag, an intent tag that matches the target text is determined.
2. The method of claim 1, further comprising:
inputting the label elements into a pre-constructed first identification model, and determining predicted intention labels corresponding to the label elements through the first identification model, wherein the first identification model has the capacity of enabling the intention labels corresponding to the label elements to trend to actual intention labels corresponding to the label elements;
and creating an intention database according to the label elements and the intention labels matched with the label elements.
3. The method of claim 2, further comprising:
obtaining sample information, the sample information including a label element and an intention label;
and respectively taking the obtained sample information as the training input of the neural network, and training to obtain a first recognition model.
4. The method of claim 1, the determining an intent tag that matches the target text based on the first intent tag, comprising:
performing intention identification on the target text to obtain a second intention label;
and performing fusion processing on the first intention label and the second intention label to obtain an intention label of the target text.
5. The method of claim 4, wherein the identifying the intent of the target text to obtain a second intent tag comprises:
inputting the target text into a pre-constructed second recognition model, and determining a predicted second intention label corresponding to the target text through the second recognition model;
wherein the second recognition model has the capability of trending a second intention label corresponding to the target text towards an actual intention label corresponding to the target text; and the second recognition model is a model obtained by training by taking the obtained sample information as the training input of the neural network, wherein the sample information is matched with the target text information.
6. The method of claim 1, wherein the splitting the target text into tag elements to obtain tag elements comprises:
and extracting the features of the target text, and determining the label elements according to the extracted features.
7. The method of claim 1, wherein the splitting the target text into tag elements to obtain tag elements comprises:
classifying the target text by utilizing a pre-established classification model to obtain a label element; and the pre-created classification model representation judges whether the target text contains a model of a specific tag element.
8. An information processing apparatus comprising:
an acquisition unit configured to acquire a target text;
the splitting unit is used for splitting the label elements of the target text to obtain the label elements, and the label elements represent key words with specific semantics;
the query unit is used for querying in an intention database by using the tag elements to obtain a first intention tag matched with the tag elements, the intention database comprises a tag element set and an intention tag set, and the tag element set and the intention tag set have a preset mapping relation;
a determining unit, configured to determine, based on the first intention tag, an intention tag that matches the target text.
9. An electronic device, comprising:
a memory for storing a program;
a processor configured to execute the program, the program specifically configured to:
acquiring a target text;
splitting the target text to obtain tag elements, wherein the tag elements represent key words with specific semantics;
querying in an intention database by using the label element to obtain a first intention label matched with the label element, wherein the intention database comprises a label element set and an intention label set, and the label element set and the intention label set have a preset mapping relation;
based on the first intent tag, an intent tag that matches the target text is determined.
10. A storage medium characterized in that the storage medium stores computer program code which, when executed, implements an information processing method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911316145.1A CN111078855A (en) | 2019-12-19 | 2019-12-19 | Information processing method, information processing device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911316145.1A CN111078855A (en) | 2019-12-19 | 2019-12-19 | Information processing method, information processing device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111078855A true CN111078855A (en) | 2020-04-28 |
Family
ID=70315605
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911316145.1A Pending CN111078855A (en) | 2019-12-19 | 2019-12-19 | Information processing method, information processing device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111078855A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112905757A (en) * | 2021-01-27 | 2021-06-04 | 北京金山数字娱乐科技有限公司 | Text processing method and device |
| CN113593533A (en) * | 2021-09-10 | 2021-11-02 | 平安科技（深圳）有限公司 | Flow node skipping method, device, equipment and medium based on intention recognition |
| CN115910110A (en) * | 2022-11-30 | 2023-04-04 | 杭州网筌科技有限公司 | Government affair service system based on natural language identification |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688614A (en) * | 2017-08-04 | 2018-02-13 | 平安科技（深圳）有限公司 | It is intended to acquisition methods, electronic installation and computer-readable recording medium |
| CN110096570A (en) * | 2019-04-09 | 2019-08-06 | 苏宁易购集团股份有限公司 | A kind of intension recognizing method and device applied to intelligent customer service robot |
| CN110162633A (en) * | 2019-05-21 | 2019-08-23 | 深圳市珍爱云信息技术有限公司 | Voice data is intended to determine method, apparatus, computer equipment and storage medium |
| CN110276067A (en) * | 2019-05-07 | 2019-09-24 | 阿里巴巴集团控股有限公司 | Text is intended to determine method and device |
| CN110399609A (en) * | 2019-06-25 | 2019-11-01 | 众安信息技术服务有限公司 | Intension recognizing method, device, equipment and computer readable storage medium |
| CN110580335A (en) * | 2018-06-07 | 2019-12-17 | 阿里巴巴集团控股有限公司 | user intention determination method and device |
-
2019
- 2019-12-19 CN CN201911316145.1A patent/CN111078855A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688614A (en) * | 2017-08-04 | 2018-02-13 | 平安科技（深圳）有限公司 | It is intended to acquisition methods, electronic installation and computer-readable recording medium |
| CN110580335A (en) * | 2018-06-07 | 2019-12-17 | 阿里巴巴集团控股有限公司 | user intention determination method and device |
| CN110096570A (en) * | 2019-04-09 | 2019-08-06 | 苏宁易购集团股份有限公司 | A kind of intension recognizing method and device applied to intelligent customer service robot |
| CN110276067A (en) * | 2019-05-07 | 2019-09-24 | 阿里巴巴集团控股有限公司 | Text is intended to determine method and device |
| CN110162633A (en) * | 2019-05-21 | 2019-08-23 | 深圳市珍爱云信息技术有限公司 | Voice data is intended to determine method, apparatus, computer equipment and storage medium |
| CN110399609A (en) * | 2019-06-25 | 2019-11-01 | 众安信息技术服务有限公司 | Intension recognizing method, device, equipment and computer readable storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112905757A (en) * | 2021-01-27 | 2021-06-04 | 北京金山数字娱乐科技有限公司 | Text processing method and device |
| CN113593533A (en) * | 2021-09-10 | 2021-11-02 | 平安科技（深圳）有限公司 | Flow node skipping method, device, equipment and medium based on intention recognition |
| WO2023035524A1 (en) * | 2021-09-10 | 2023-03-16 | 平安科技（深圳）有限公司 | Intention recognition-based process node jump method and apparatus, device, and medium |
| CN113593533B (en) * | 2021-09-10 | 2023-05-02 | 平安科技（深圳）有限公司 | Method, device, equipment and medium for jumping flow node based on intention recognition |
| CN115910110A (en) * | 2022-11-30 | 2023-04-04 | 杭州网筌科技有限公司 | Government affair service system based on natural language identification |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107491534B (en) | Information processing method and device | |
| US11151177B2 (en) | Search method and apparatus based on artificial intelligence | |
| CN111026842B (en) | Natural language processing method, natural language processing device and intelligent question-answering system | |
| US20230237328A1 (en) | Information processing method and terminal, and computer storage medium | |
| CN116775847B (en) | Question answering method and system based on knowledge graph and large language model | |
| US11163936B2 (en) | Interactive virtual conversation interface systems and methods | |
| CN111428010B (en) | Man-machine intelligent question-answering method and device | |
| US11636936B2 (en) | Method and apparatus for verifying medical fact | |
| US11238132B2 (en) | Method and system for using existing models in connection with new model development | |
| CN111078855A (en) | Information processing method, information processing device, electronic equipment and storage medium | |
| CN110929125A (en) | Search recall method, apparatus, device and storage medium thereof | |
| CN112035638B (en) | Information processing method, device, storage medium and equipment | |
| US20200380311A1 (en) | Collaborative Information Extraction | |
| CN113342944A (en) | Corpus generalization method, apparatus, device and storage medium | |
| CN112417996A (en) | Information processing method and device for industrial drawing, electronic equipment and storage medium | |
| CN116701604A (en) | Question and answer corpus construction method and device, question and answer method, equipment and medium | |
| CN114398903B (en) | Intention recognition method, device, electronic equipment and storage medium | |
| CN114528851B (en) | Reply sentence determination method, reply sentence determination device, electronic equipment and storage medium | |
| CN111597453B (en) | User image drawing method, device, computer equipment and computer readable storage medium | |
| CN113946668A (en) | Semantic processing method, system and device based on edge node and storage medium | |
| CN113627161A (en) | Data processing method and device, storage medium and electronic equipment | |
| CN112559697A (en) | Method and device for recognizing conversational intention | |
| CN117473070B (en) | Multi-channel application method of intelligent robot, intelligent robot and storage medium | |
| CN117648986B (en) | Task processing and code processing method, computing device, medium, and program product | |
| US20240202228A1 (en) | Systems and methods for dynamically generating groups of receveid textual data for collective labeling |
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 |