CN115599899A - Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph - Google Patents
Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph Download PDFInfo
- Publication number
- CN115599899A CN115599899A CN202211388973.8A CN202211388973A CN115599899A CN 115599899 A CN115599899 A CN 115599899A CN 202211388973 A CN202211388973 A CN 202211388973A CN 115599899 A CN115599899 A CN 115599899A
- Authority
- CN
- China
- Prior art keywords
- entity
- question
- aircraft
- knowledge graph
- attribute
- 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
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
- G06F16/33—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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to the technical field of artificial intelligence question answering, and discloses an intelligent question answering method, system, equipment and medium based on an aircraft knowledge graph, wherein the method comprises the following steps: performing semantic analysis on the user problem to generate query data; retrieving a triple corresponding to the query data from a pre-constructed aircraft knowledge graph; outputting answers to the questions according to the triples obtained by searching; and aiming at the satisfaction evaluation of the user on the answer to the question, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge map library, and updating the map library and the question-answering model in real time. Therefore, intelligent question answering and online updating are completed through four steps of semantic analysis preprocessing, triple retrieval, question answer output and knowledge updating, retrieved information in the process is derived from a knowledge map database, one-time construction can be achieved, subsequent information can be rapidly retrieved and inquired, the calculated amount in the subsequent question answering process is reduced, and the speed and the efficiency of a question answering system are improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligent question answering, in particular to an intelligent question answering method, system, equipment and medium based on an aircraft knowledge graph.
Background
In the aerospace field related to the aircraft, due to the complexity and the specialty of the disciplines, a great technical barrier exists for non-professionals, and certain difficulties and obstacles exist in acquiring professional knowledge of the relevant aircraft. As a tool for realizing the autonomous interactive question answering between a human and a machine, the intelligent question answering system can provide information services such as real-time information retrieval, deep data mining and the like, and is becoming a mainstream information acquisition mode.
At present, the traditional question-answering model based on unstructured free text reading understanding mainly has three problems: the method has the advantages that firstly, the aircraft data are widely distributed, the multi-source data of various websites, books and literature data need to be matched, extracted, fused and analyzed on line, the data volume is large, the processing speed is low, and the question and answer efficiency is influenced; secondly, aircraft part professional terms often have longer explanatory languages, and a traditional model based on a regular matching method can only match text characters according to specific rules and cannot match text meanings, so that the processing effect on situations such as polysemous words and similar words is poor, and the answer precision is lower although the method is simpler; thirdly, for a complex problem with a plurality of candidate answers, such as the problem of 'the oil loading amount of an F-35 fighter', the different models of F-35 have different oil loading amounts, so that the complex problem needs to answer one by one for each model, and a traditional question-answer model usually screens out a final answer by simply sequencing the candidate answers, but does not fully consider the correlation information among the candidate answers, so that the obtained answer may not be comprehensive enough.
Therefore, how to solve the problems of low efficiency, low precision and the like of the existing question-answering model is a technical problem to be urgently solved by the technical personnel in the field.
Disclosure of Invention
In view of this, the present invention provides an intelligent question-answering method, system, device and medium based on an aircraft knowledge graph, which can reduce the amount of calculation in the question-answering process, improve the reaction speed and efficiency, and improve the accuracy of the obtained answers. The specific scheme is as follows:
an intelligent question-answering method based on an aircraft knowledge graph comprises the following steps:
performing semantic analysis on the user problem to generate query data;
searching and obtaining triples corresponding to the query data in a pre-constructed aircraft knowledge graph;
outputting answers to the questions according to the triples obtained by searching;
and aiming at the satisfaction evaluation of the user on the answer of the question or the new entity in the user question, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge atlas library, and updating the atlas library and the question-answer model in real time.
Preferably, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, the performing semantic analysis on the user question to generate query data includes:
performing word segmentation and entity recognition on the user question to obtain a corresponding entity recognition result;
judging the problem type according to the entity recognition result and the keyword matching method to obtain the corresponding problem type;
and generating corresponding query data according to the entity identification result and the obtained problem type.
Preferably, in the above intelligent question-answering method based on an aircraft knowledge graph provided by the embodiment of the present invention, the performing word segmentation and entity recognition on the user question includes:
performing word segmentation and semantic coding on the user problem by adopting a pre-trained BERT model;
and performing depth decoding by adopting a BilSTM-CRF model to complete entity identification.
Preferably, in the above intelligent question-answering method based on the aircraft knowledge graph provided by the embodiment of the invention, the problem type judgment is carried out according to the entity recognition result and the keyword matching method to obtain the corresponding problem type, and the method comprises the following steps:
obtaining the number of the identified entities from the entity identification result, and judging the problem type to be a single entity problem or a multi-entity problem;
when the problem is a single entity problem, obtaining the number of attribute words and the existing keywords in the user problem from the entity recognition result, and judging the problem type to be a single entity multi-attribute problem, a single entity single-attribute problem or a single entity non-generic problem through a keyword matching method;
and when the problem is a multi-entity problem, obtaining the number of attribute words and existing keywords in the user problem from the entity identification result, and judging the problem types to be a multi-entity multi-attribute problem, a multi-entity single-attribute problem, a multi-entity non-category problem or a multi-entity comparison category problem by a keyword matching method.
Preferably, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, the retrieving, in the pre-constructed aircraft knowledge graph, a triple corresponding to the query data includes:
searching the same entity words in a triple database of a pre-constructed aircraft knowledge graph according to the entity words in the query data, and linking the same entity words to correct entities;
comparing and matching the entity attribute words in the query data with the pre-constructed template attribute words; and if the matching is successful, returning the entity attribute words in the triple database of the aircraft knowledge graph corresponding to the template attribute words.
Preferably, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, after the comparing and matching of the entity attribute words in the query data with the pre-constructed template attribute words, the method further includes:
if the matching fails, establishing an entity and an entity attribute link by adopting a Manhattan LSTM model; the Manhattan LSTM model takes a hidden layer as a final layer, and takes the state of the last time sequence of the hidden layer as an input parameter of similarity coefficient calculation.
Preferably, in the above intelligent question-answering method based on an aircraft knowledge graph provided in an embodiment of the present invention, the outputting answers to questions according to the triples obtained by retrieval includes:
extracting attribute values of the triples according to the triples obtained by searching;
and combining the extracted attribute values to obtain the question type, combining the question answers and outputting the question answers.
The embodiment of the invention also provides an intelligent question-answering system based on the aircraft knowledge graph, which comprises:
the database module is used for storing the pre-constructed aircraft knowledge graph on a storage medium;
the back-end processing module is used for executing intelligent question answering; the back-end processing module comprises a preprocessing unit, a retrieval unit, an output unit and an updating unit; wherein the content of the first and second substances,
the preprocessing unit is used for performing semantic analysis on the user problem to generate query data;
the retrieval unit is used for retrieving and obtaining the triples corresponding to the query data in the aircraft knowledge graph;
the output unit is used for outputting answers to the questions according to the triples obtained by retrieval;
the updating unit is used for adding new attribute words or entities which may appear into a system template library or an aircraft knowledge map library according to the satisfaction evaluation of the user on the answer of the question or new entities in the user question, and updating the map library and the question and answer model in real time;
and the front-end display module is used for displaying the user question and the question answer output by the back-end processing module.
The embodiment of the invention also provides intelligent question-answering equipment based on the aircraft knowledge graph, which comprises a processor and a memory, wherein the processor is used for realizing the intelligent question-answering method based on the aircraft knowledge graph provided by the embodiment of the invention when executing the computer program stored in the memory.
The embodiment of the present invention further provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned intelligent question-answering method based on the aircraft knowledge graph provided by the embodiment of the present invention.
According to the technical scheme, the intelligent question-answering method based on the aircraft knowledge graph comprises the following steps: performing semantic analysis on the user problem to generate query data; retrieving a triple corresponding to the query data from a pre-constructed aircraft knowledge graph; outputting answers to the questions according to the triples obtained by searching; and aiming at the satisfaction evaluation of the user on the answer of the question or the new entity in the user question, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge atlas library, and updating the atlas library and the question-answer model in real time.
According to the intelligent question-answering method based on the aircraft knowledge graph, provided by the invention, intelligent question-answering and online updating are completed through four steps of semantic analysis preprocessing, triple retrieval, question answer output and knowledge updating, retrieved information in the process is derived from a knowledge graph database, one-time construction can be realized, the subsequent information can be quickly and accurately retrieved and inquired, the calculated amount in the subsequent question-answering process is reduced, the speed and the efficiency of a question-answering system are greatly improved, and the accuracy of the obtained answer is higher.
In addition, the invention also provides a corresponding system, equipment and a computer readable storage medium aiming at the intelligent question-answering method based on the aircraft knowledge graph, so that the method has higher practicability, and the system, the equipment and the computer readable storage medium have corresponding advantages.
Drawings
In order to more clearly illustrate the embodiments of the present invention or technical solutions in related arts, the drawings used in the description of the embodiments or related arts will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of an intelligent question-answering method based on an aircraft knowledge graph according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a framework of steps in an intelligent question-answering method based on an aircraft knowledge graph according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of entity identification by combining the BERT model with the BilSTM-CRF model according to the embodiment of the present invention;
FIG. 4 is a schematic diagram of a computing process of the Manhattan LSTM model according to the embodiment of the present invention;
fig. 5 is a schematic structural diagram of an intelligent question-answering system based on an aircraft knowledge graph according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, 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 invention.
The invention provides an intelligent question-answering method based on an aircraft knowledge graph, which comprises the following steps as shown in figure 1:
s101, performing semantic analysis on a user problem to generate query data;
it should be noted that step S101 is a preprocessing stage for the user question. Since the problem posed by the user usually occurs in the form of a natural language, it needs to be converted into a formalized language that can be understood by a computer, i.e., the required query data. This step may be performed by a pre-processing unit to perform semantic analysis of the user's question and generate query data.
S102, retrieving and obtaining triples corresponding to query data in a pre-constructed aircraft knowledge graph;
it will be appreciated that a knowledge graph is essentially a semantic network, containing entities and entity relationships/attributes. Entity relationships represent entities and associations between entities, and attributes in the knowledge graph represent certain characteristics that an entity has which describe the relevant information of the entity. The primary representation of a knowledge graph is a triplet, including an attribute triplet and a relationship triplet. Step S102 is a stage of retrieving triples in the knowledge graph, where the retrieving unit is mainly responsible for receiving query data, and retrieving matched triples in the pre-constructed knowledge graph according to the query data, i.e. an entity linking process.
And S103, outputting answers to the questions according to the triples obtained by searching.
Specifically, step S103 is a question answer output stage, and the output unit may obtain and output the question answer by extracting the triple attribute information.
And S104, aiming at the satisfaction evaluation of the user on the answer of the question or the new entity in the question of the user, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge atlas library, and updating the atlas library and the question-answer model in real time.
It should be noted that, with the continuous operation of the aircraft question-answering system, the accumulated professional question-answering knowledge will be more abundant, the scale of the question-answering records stored therein will be increased, and the professional domain knowledge base can be updated online by using the obtained question-answering records. Specifically, the question-answer satisfaction evaluation is set at the end of each question-answer, the user can score the accuracy of the answer, and if the score is higher than the set acceptance value, the question attribute words obtained in the process are added into the template attribute words (if the attribute words do not appear in the template attribute words yet), so that the efficiency and accuracy of the subsequent question-answer are improved. If a new entity appears in the problem, such as the new aircraft is successfully developed, the new entity appearing in the problem is added into the map library, the new entity and the existing attribute words are automatically associated according to knowledge reasoning and completion, and meanwhile, the corresponding attribute values are extracted from the public data by using technologies such as crawler and the like to construct a new map node. When the question-answering system is used, the new database is used for model retraining according to the updating condition of the map database, for example, the accumulated updated entry of the database exceeds a set value, so that the question-answering model can be updated in real time, and the answering speed and accuracy of the system are improved.
In the intelligent question-answering method based on the aircraft knowledge graph provided by the embodiment of the invention, intelligent question-answering and online updating are completed through four steps of semantic analysis preprocessing, triple retrieval, question answer output and knowledge updating, the process is simple, the retrieved information is derived from a knowledge graph database, one-time construction can be realized, the subsequent information can be quickly and accurately retrieved and inquired, the calculated amount in the subsequent question-answering process is reduced, the speed and the efficiency of a question-answering system are greatly improved, and the accuracy of the obtained answer is higher.
In practical application, the intelligent question-answering method based on the aircraft knowledge graph has important practical significance for realizing that various practitioners in the aerospace field can conveniently and quickly acquire panoramic information of related aircraft, reducing information barriers and promoting and popularizing aerospace vehicle knowledge of people, and can provide beneficial reference for the construction of intelligent question-answering systems in other vertical fields.
Further, in specific implementation, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, the step S101 performs semantic analysis on the user question to generate query data, which may include:
firstly, performing word segmentation and entity recognition on a user question to obtain a corresponding entity recognition result; the entity recognition result may include entities, entity attribute words, the number of entities, and the like;
then, judging the problem type according to the entity recognition result and the keyword matching method to obtain the corresponding problem type;
and finally, generating corresponding query data according to the entity recognition result and the obtained problem type.
As shown in fig. 2, step S101 mainly completes the word segmentation, entity identification and problem type determination, and finally obtains the word segmentation sequence of the user problem, extracts the entities and entity attribute words in the user problem, and generates query data according to the problem type. Taking the problem "how many the production quantity of the bird fighters are" as an example, the user problem needs to be participled and entity identification is performed to identify the entity "bird fighter" and the entity attribute "production quantity" in the user problem, so as to generate the required query data.
In specific implementation, the performing word segmentation and entity recognition on the user question in the above steps may include: performing word segmentation and semantic coding on the user problem by adopting a pre-trained BERT (Bidirectional Encoder descriptions from Transformer) model; and performing depth decoding by adopting a BilSTM-CRF model to complete entity identification.
As shown in fig. 3, the concrete implementation of the BERT model in entity recognition is mainly divided into two steps, namely pre-training (pre-training) and fine-tuning (fine-tuning). The pre-training is equivalent to word embedding, a model is trained by utilizing the corpus without any mark, and a BertChinese pre-training model is adopted in the invention. The fine tuning process utilizes the trained model to set specific input and output portions based on different tasks (e.g., entity recognition in the present invention) to accomplish related tasks. When entity recognition is carried out, the fine tuning process of the BERT model is mainly used for carrying out the Encoder process of semantic coding, the traditional word embedding step is replaced, and because the fine tuning process is based on the pre-trained model, a smaller data set can be adopted, training can be rapidly completed, and good performance and results are obtained. In some cases, better entity recognition results can be obtained after fine tuning. In order to achieve a better effect, the BilSTM model and the CRF model are used for deep decoding after the Bert so as to improve the accuracy of entity identification. By adopting the artificial intelligence method, the accuracy of word segmentation and entity recognition can be improved. Among them, bilSTM (Bidirectional Long-Short Term)Memory) is a bidirectional long-time Memory network, a long association relation (which can embody context information in an aircraft text) can be stored by means of a storage unit structure of a model, the Memory network can be divided into a forward LSTM and a backward LSTM according to the time sequence of hidden layer state transmission, and the corresponding hidden layer states are respectively a forward LSTM and a backward LSTM
、
(k =1,2,3,4 …). The CRF (Conditional Random Fields) layer is a Conditional Random field, and can consider entity sequence labels at a statement level. The combination of BilSTM and CRF can ensure that the entity sequence labeling can be effectively carried out while extracting enough context information, and further entity labels, such as B (representing the beginning of an entity keyword), O (representing a non-entity keyword) and I (representing the non-first character of an entity), can be obtained.
In specific implementation, the step of determining the problem type according to the entity recognition result and the keyword matching method to obtain a corresponding problem type may include: obtaining the number of the identified entities from the entity identification result, and judging the problem type to be a single entity problem or a multi-entity problem; when the problem is a single-entity problem, the number of attribute words and existing keywords in the user problem are obtained from the entity recognition result, and the problem type is judged to be a single-entity multi-attribute problem, a single-entity single-attribute problem or a single-entity non-category problem through a keyword matching method; and when the problem is a multi-entity problem, obtaining the number of attribute words and the existing keywords in the user problem from the entity recognition result, and judging the problem types to be a multi-entity multi-attribute problem, a multi-entity single-attribute problem, a multi-entity non-class problem or a multi-entity comparison problem by a keyword matching method.
Firstly, judging the problem category according to the entity quantity obtained by the entity identification result, and if the identified entity quantity is one, determining the obtained problem type to be a single entity problem; if the number of identified entities exceeds one, the resulting problem type is a multi-entity problem. Secondly, a keyword matching method is adopted to judge the specific problem types, namely, keywords for marking the problem types are searched in the question sentences. If a plurality of attribute words exist in the user problem and the existing keywords can comprise 'how many' or 'what', the obtained problem type is a single-entity multi-attribute problem or a multi-entity multi-attribute problem; if a single attribute word exists in the user problem and the existing keywords can include 'how many' or 'what', the obtained problem type is a single-entity single-attribute problem or a multi-entity single-attribute problem; if a single attribute word exists in the user problem and an attribute value exists at the same time, and/or the existing keyword can comprise 'do', the obtained problem type is that a single entity is a non-class problem or multiple entities are non-class problems; if multiple entities and a single attribute word exist in the user question at the same time, and the existing keywords can include "than" or "more", the obtained question type is a multiple-entity comparison class.
Specifically, the problem type is judged after the entity identification result is obtained. The common user problems can be divided into seven types according to the number of entities and the types of the problems, wherein the seven types are respectively single entity single attribute, single entity multiple attribute, single entity non-class, multiple entity single attribute, multiple entity multiple attribute, multiple entity non-class and multiple entity comparison class, as shown in table one.
TABLE-aircraft knowledge problem type Classification
And finally, generating corresponding query data according to the entity identification result and the problem type. If the problem is a single entity single attribute type problem, the entity and the entity attribute word are used as query data; and if the single entity is a non-generic problem, taking the entity, the entity attribute word and the attribute value as query data. If the problem is a multi-entity comparison problem, a plurality of entities and corresponding entity attribute words are required to be used as query data.
Further, in a specific implementation, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, the step S102 of retrieving a triple corresponding to query data from a pre-constructed aircraft knowledge graph may include: searching the same entity words in a triple database of a pre-constructed aircraft knowledge graph according to the entity words in the query data, and linking the entity words to correct entities; comparing and matching entity attribute words in query data with pre-constructed template attribute words; and if the matching is successful, returning the entity attribute words in the triple database of the aircraft knowledge graph corresponding to the template attribute words. Therefore, the extraction, fusion and processing of multi-source information can be realized, and the precision of the question-answering system is further obviously improved.
Specifically, in the above process, a method of combining matching and deep learning is used. Firstly, adopting a matching method to link, searching the same entity words in a triple database of a knowledge graph according to the entity words generated in a query stage, and linking to correct entities; and then comparing and matching the submitted entity attribute words with the pre-constructed template attribute words, and if the entity attribute words in the attribute query data exist in the template attribute words, returning the corresponding entity attribute words of the knowledge graph library. Wherein the template attribute words are constructed manually according to the corpus dictionary. And part of the template attribute words are shown in a table two.
Template word matching template for question in table two parts
Although the template word matching method is high in speed and accuracy, the user problem is high in uncertainty, and the manually defined template word bank cannot be matched with the real user problem, for example, "what is the main use scene of F-35", the problem is similar to the type of inquiring F-35, but the template word matching method is difficult to use. Because the uncertainty of the user question is difficult to exhaust all template words, in specific implementation, in the above intelligent question-answering method based on the aircraft knowledge graph provided by the embodiment of the present invention, after comparing and matching the entity attribute words in the query data with the pre-constructed template attribute words in the step described above, the method may further include: if matching fails, namely a matching word is not found in the template word bank, establishing entity and entity attribute link by adopting a Manhattan LSTM (MaLSTM for short) model so as to ensure the robustness of the question-answering system.
Conventional LSTM includes a hidden layer
And a cell layer
The input layer sequentially passes through the hidden layer and the cell layer, and the implementation process of the MalsTM is slightly different, and the hidden layer is taken as the final layer through the following calculation steps:
wherein the content of the first and second substances,
，
representing time of daytThe hidden layer state of the location is,
indicating the time of daytThe left-behind door of the automobile is opened,σa sigmoid function is represented as a function,
a weight matrix representing a forgetting gate,x t indicating the input state of the network at the current time,b f a bias term representing a forgetting gate,i t indicating the time of daytThe input gate of (a) the input gate,
a weight matrix representing the input gates is shown,b i presentation inputThe biasing term of the door is such that,c t indicating the state of the cell at the current time,
indicating the memory state at the present moment,
a weight matrix representing a memory state,b c a bias term representing a memory state,o t indicating the time of daytThe output gate of (a) the output gate of (b),W 0 a weight matrix representing the output gates,b o representing the bias term of the output gate. Obtaining a hidden layer
Then, the state of the last time series is taken
Input parameters calculated for the similarity coefficients.
FIG. 4 shows a schematic of the computational flow of the MalSTM model. The whole processing process is divided into a network LSTM (a) and LSTM (b) of a and b. Taking LSTM (a) network as an example, firstly, the word vector after the words of the question are divided into words
(k =1,2,3, … n) as an input layer is transmitted to the LSTM network a, and the hidden layer state at each time is obtained
(k =1,2,3,. N); the network LSTM (b) is processed in the same way, and the triples in the knowledge graph are input as word vectors, namely
(k =1,2,3, … n) as input, and sends it to the LSTM (b) network to obtain the hidden layer state at each time
(k =1,2,3,. N). Finally, the last time in the two networksnHidden layer state of
、
Extracted and then using a formula
Obtaining a similarity coefficient
Wherein
. And sequencing the obtained similarity coefficients, taking the knowledge graph triple with the highest value, comparing the value with a set threshold, and if the value is greater than or equal to the threshold, establishing entity and entity attribute links, otherwise, not establishing the link.
For comparison problems, corresponding triples need to be found, linked to multiple graph entities.
In this way, the entity link is carried out by adopting the template matching and the similarity calculation technology based on deep learning, the robustness of the system is improved while the question-answering efficiency is ensured, meanwhile, the system can supplement a question-answering database in real time, the accurate answer to some special problems is realized, and the question-answering satisfaction degree is improved.
Further, in a specific implementation, in the above intelligent question-answering method based on an aircraft knowledge graph provided in the embodiment of the present invention, as shown in fig. 2, the step S103 outputs answers to the questions according to the retrieved triples, which may include: extracting attribute information of the triples according to the triples obtained by searching; and combining the extracted attribute information with the obtained question types, combining the question answers and outputting the question answers.
Specifically, in step S103, attributes of the ternary groups are extracted from the ternary groups obtained by the search, and answers are combined and output according to the question types. If the question ' what the code number of the F-35 is ' and the triple is ' F-35-alias-bird ', the answer ' the code number of the F-35 is ' bird ' is generated by combining the entity and the attribute word in the question and the attribute word in the triple and adding the auxiliary word, and the answer is output to the user terminal.
Based on the same inventive concept, the embodiment of the invention also provides an intelligent question-answering system based on the aircraft knowledge graph, and as the problem solving principle of the system is similar to that of the intelligent question-answering method based on the aircraft knowledge graph, the implementation of the system can refer to the implementation of the intelligent question-answering method based on the aircraft knowledge graph, and repeated parts are not repeated.
In specific implementation, the intelligent question-answering system based on the aircraft knowledge graph provided by the embodiment of the invention, as shown in fig. 5, may specifically include:
the database module 1 is used for storing a pre-constructed aircraft knowledge graph on a storage medium; specifically, the aircraft knowledge graph is based on a Resource Description Framework (RDF) and is stored on a storage medium in an SQL database format, and the storage medium has a processor matched with the storage medium to execute instructions;
the back-end processing module 2 is used for executing the whole intelligent question answering and can be written by adopting object-oriented languages such as python; the back-end processing module may include a preprocessing unit 21, a retrieving unit 22, an output unit 23, and an updating unit 24; wherein, the first and the second end of the pipe are connected with each other,
the preprocessing unit 21 is used for performing semantic analysis on the user problem to generate query data;
the retrieval unit 22 is configured to retrieve the triples corresponding to the query data from the aircraft knowledge graph;
the output unit 23 is configured to output answers to the questions according to the triples obtained by the retrieval;
the updating unit 24 is configured to add new attribute words or entities that may appear into the system template base or the aircraft knowledge atlas base according to the satisfaction evaluation of the user on the answer to the question or new entities in the user question, and update the atlas base and the question-and-answer model in real time;
and the front-end display module 3 is used for displaying the user question and the question answer output by the back-end processing module. The front-end display module 3 can be specifically responsible for system visualization, including user input of questions and answers obtained by displaying back-end processing, can be written in languages such as javascript, and displayed by using a webpage, and adopts a mainstream micro-service design framework, so that the front-end module and the back-end module are loosely coupled, and subsequent maintenance, expansion and updating are facilitated.
In the intelligent question-answering system based on the aircraft knowledge graph provided by the embodiment of the invention, intelligent question-answering and online updating can be completed through the mutual combination of the three modules, the process is simple, the retrieved information is derived from the knowledge graph database, one-time construction can be realized, the subsequent information can be quickly and accurately retrieved and queried, the calculated amount in the subsequent question-answering process is reduced, the speed and the efficiency of the question-answering system are greatly improved, and the accuracy of the obtained answer is higher.
In specific implementation, in the above intelligent question-answering system based on the aircraft knowledge graph provided in the embodiment of the present invention, the preprocessing unit 21 is specifically configured to perform word segmentation and entity recognition on a user question to obtain a corresponding entity recognition result; judging the problem type according to the entity recognition result and the keyword matching method to obtain the corresponding problem type; and generating corresponding query data according to the entity recognition result and the obtained problem type.
In specific implementation, in the above intelligent question-answering system based on an aircraft knowledge graph provided in the embodiment of the present invention, the retrieval unit 22 is specifically configured to search for the same entity word in a pre-constructed triple database of the aircraft knowledge graph according to the entity word in the query data, and link to a correct entity; comparing and matching entity attribute words in the query data with pre-constructed template attribute words; if the matching is successful, returning the entity attribute words in the triple database of the aircraft knowledge graph corresponding to the template attribute words; if the matching fails, establishing an entity and an entity attribute link by adopting a Manhattan LSTM model; the Manhattan LSTM is an input parameter which takes a hidden layer as a final layer and takes the state of the last time sequence of the hidden layer as a similarity coefficient for calculation.
Along with the continuous operation of the aircraft question-answering system, the accumulated professional question-answering knowledge is more abundant, the scale of the question-answering records stored in the aircraft question-answering system is also improved, the professional field knowledge base is continuously updated, and the training model of the intelligent question-answering can be updated in real time.
In specific implementation, in the intelligent question-answering system based on the aircraft knowledge graph provided in the embodiment of the present invention, the output unit 23 is specifically configured to extract attribute values of the triples according to the triples obtained through retrieval; and combining the extracted attribute values with the obtained question types, combining the question answers and outputting the question answers.
For more specific working processes of the modules, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
Correspondingly, the embodiment of the invention also discloses intelligent question answering equipment based on the aircraft knowledge graph, which comprises a processor and a memory; the processor implements the intelligent question-answering method based on the aircraft knowledge graph disclosed in the previous embodiment when executing the computer program stored in the memory.
For more specific processes of the above method, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
Further, the present invention also discloses a computer readable storage medium for storing a computer program; the computer program when executed by the processor implements the intelligent question-answering method based on the aircraft knowledge graph disclosed in the foregoing.
For more specific processes of the above method, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The system, the device and the storage medium disclosed by the embodiment correspond 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.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
To sum up, the embodiment of the invention provides an intelligent question-answering method based on an aircraft knowledge graph, which comprises the following steps: performing semantic analysis on the user problem to generate query data; retrieving a triple corresponding to the query data from a pre-constructed aircraft knowledge graph; outputting answers to the questions according to the triples obtained by searching; and aiming at the satisfaction evaluation of the user on the answer of the question or the new entity in the user question, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge atlas library, and updating the atlas library and the question-answer model in real time. The intelligent question and answer method completes intelligent question and answer and online update through four steps of semantic analysis preprocessing, triple retrieval, question answer output and knowledge update, the process is simple, retrieved information is derived from a knowledge map database, one-time construction can be realized, subsequent information can be quickly and accurately retrieved and inquired, the calculated amount in the subsequent question and answer process is reduced, the reaction speed and efficiency are greatly improved, and the accuracy of the obtained answer is higher. In addition, the invention also provides a corresponding system, equipment and a computer readable storage medium aiming at the intelligent question answering method, so that the method has higher practicability, and the system, the equipment and the computer readable storage medium have corresponding advantages.
Finally, it should also be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of another identical element in a process, method, article, or apparatus that comprises the element.
The above detailed description is given to the intelligent question-answering method, system, device and medium based on the aircraft knowledge graph, and the specific examples are applied in the text to explain the principle and implementation of the invention, and the description of the above embodiments is only used to help understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. An intelligent question-answering method based on an aircraft knowledge graph is characterized by comprising the following steps:
performing semantic analysis on the user problem to generate query data;
retrieving and acquiring a triple corresponding to the query data in a pre-constructed aircraft knowledge graph;
outputting answers to the questions according to the triples obtained by searching;
and aiming at the satisfaction evaluation of the user on the answer of the question or the new entity in the user question, adding the new attribute words or entities which may appear into a system template library or an aircraft knowledge atlas library, and updating the atlas library and the question-answer model in real time.
2. The intelligent question-answering method based on the aircraft knowledge-graph according to claim 1, wherein the semantic analysis is performed on the user questions to generate query data, and the query data comprises the following steps:
performing word segmentation and entity recognition on the user question to obtain a corresponding entity recognition result;
judging the problem type according to the entity recognition result and the keyword matching method to obtain the corresponding problem type;
and generating corresponding query data according to the entity identification result and the obtained problem type.
3. The intelligent question-answering method based on the aircraft knowledge graph according to claim 2, wherein the word segmentation and entity recognition of the user questions comprises:
performing word segmentation and semantic coding on the user problem by adopting a pre-trained BERT model;
and performing deep decoding by using a BilSTM-CRF model to complete entity identification.
4. The intelligent question-answering method based on the aircraft knowledge graph according to claim 3, wherein the step of judging the question type according to the entity recognition result and the keyword matching method to obtain the corresponding question type comprises the following steps:
obtaining the number of the identified entities from the entity identification result, and judging the problem type to be a single entity problem or a multi-entity problem;
when the problem is a single entity problem, obtaining the number of attribute words and existing keywords in the user problem from the entity recognition result, and judging the problem type to be a single entity multi-attribute problem, a single entity single-attribute problem or a single entity non-category problem by a keyword matching method;
and when the problem is a multi-entity problem, obtaining the number of attribute words and existing keywords in the user problem from the entity identification result, and judging the problem types to be a multi-entity multi-attribute problem, a multi-entity single-attribute problem, a multi-entity non-category problem or a multi-entity comparison category problem by a keyword matching method.
5. The intelligent question-answering method based on the aircraft knowledge graph according to claim 4, wherein the retrieving of the triples corresponding to the query data in the pre-constructed aircraft knowledge graph comprises:
searching the same entity words in a triple database of a pre-constructed aircraft knowledge graph according to the entity words in the query data, and linking the same entity words to correct entities;
comparing and matching the entity attribute words in the query data with the pre-constructed template attribute words; and if the matching is successful, returning the entity attribute words in the triple database of the aircraft knowledge graph corresponding to the template attribute words.
6. The intelligent question-answering method based on the aircraft knowledge graph according to claim 5, wherein after the comparing and matching of the entity attribute words in the query data with the pre-constructed template attribute words, the method further comprises:
if the matching fails, establishing an entity and an entity attribute link by adopting a Manhattan LSTM model; the Manhattan LSTM model takes a hidden layer as a final layer, and takes the state of the last time sequence of the hidden layer as an input parameter of similarity coefficient calculation.
7. The intelligent question-answering method based on the aircraft knowledge graph according to claim 6, wherein the step of outputting answers to questions according to the triples obtained by retrieval comprises the following steps:
extracting attribute values of the triples according to the triples obtained through retrieval;
and combining the extracted attribute values to obtain the question type, combining the question answers and outputting the question answers.
8. An intelligent question-answering system based on an aircraft knowledge graph is characterized by comprising:
the database module is used for storing the pre-constructed aircraft knowledge graph on a storage medium;
the back-end processing module is used for executing intelligent question answering; the back-end processing module comprises a preprocessing unit, a retrieval unit, an output unit and an updating unit; wherein the content of the first and second substances,
the preprocessing unit is used for performing semantic analysis on the user problem to generate query data;
the retrieval unit is used for retrieving and obtaining the triple corresponding to the query data in the aircraft knowledge graph;
the output unit is used for outputting answers to the questions according to the triples obtained by retrieval;
the updating unit is used for adding new attribute words or entities which may appear into a system template library or an aircraft knowledge map library according to the satisfaction evaluation of the user on the answer of the question or new entities in the user question, and updating the map library and the question and answer model in real time;
and the front-end display module is used for displaying the user question and the question answer output by the back-end processing module.
9. An intelligent question-answering device based on an aircraft knowledge graph, which is characterized by comprising a processor and a memory, wherein the processor implements the intelligent question-answering method based on the aircraft knowledge graph according to any one of claims 1 to 7 when executing a computer program stored in the memory.
10. A computer-readable storage medium for storing a computer program, wherein the computer program when executed by a processor implements the aircraft knowledge-graph-based smart question-answering method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211388973.8A CN115599899B (en) | 2022-11-08 | 2022-11-08 | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211388973.8A CN115599899B (en) | 2022-11-08 | 2022-11-08 | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115599899A true CN115599899A (en) | 2023-01-13 |
| CN115599899B CN115599899B (en) | 2023-04-07 |
Family
ID=84852999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211388973.8A Active CN115599899B (en) | 2022-11-08 | 2022-11-08 | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115599899B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116595339A (en) * | 2023-07-19 | 2023-08-15 | 东方空间技术(山东)有限公司 | Intelligent processing method, device and equipment for space data |
| CN116738572A (en) * | 2023-06-19 | 2023-09-12 | 北京知元创通信息技术有限公司 | Aircraft configuration modularized assembly method and system |
| CN117093679A (en) * | 2023-06-19 | 2023-11-21 | ***科技（杭州）有限公司 | Large language model intelligent inquiry dialogue method, system, equipment and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919655A (en) * | 2017-01-24 | 2017-07-04 | 网易（杭州）网络有限公司 | A kind of answer provides method and apparatus |
| CN110309268A (en) * | 2019-07-12 | 2019-10-08 | 中电科大数据研究院有限公司 | A kind of cross-language information retrieval method based on concept map |
| CN110457442A (en) * | 2019-08-09 | 2019-11-15 | 国家电网有限公司 | The knowledge mapping construction method of smart grid-oriented customer service question and answer |
| CN113312500A (en) * | 2021-06-24 | 2021-08-27 | 河海大学 | Method for constructing event map for safe operation of dam |
| CN113505586A (en) * | 2021-06-07 | 2021-10-15 | 中电鸿信信息科技有限公司 | Seat-assisted question-answering method and system integrating semantic classification and knowledge graph |
| CN113535917A (en) * | 2021-06-30 | 2021-10-22 | 山东师范大学 | Intelligent question-answering method and system based on travel knowledge map |
| CN113806513A (en) * | 2021-09-30 | 2021-12-17 | 中国人民解放军国防科技大学 | Question-answering system construction method and system based on knowledge graph in military field |
| CN114416939A (en) * | 2021-12-28 | 2022-04-29 | 广州市瑞行科技有限公司 | Intelligent question and answer method, device, equipment and storage medium |
-
2022
- 2022-11-08 CN CN202211388973.8A patent/CN115599899B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106919655A (en) * | 2017-01-24 | 2017-07-04 | 网易（杭州）网络有限公司 | A kind of answer provides method and apparatus |
| CN110309268A (en) * | 2019-07-12 | 2019-10-08 | 中电科大数据研究院有限公司 | A kind of cross-language information retrieval method based on concept map |
| CN110457442A (en) * | 2019-08-09 | 2019-11-15 | 国家电网有限公司 | The knowledge mapping construction method of smart grid-oriented customer service question and answer |
| CN113505586A (en) * | 2021-06-07 | 2021-10-15 | 中电鸿信信息科技有限公司 | Seat-assisted question-answering method and system integrating semantic classification and knowledge graph |
| CN113312500A (en) * | 2021-06-24 | 2021-08-27 | 河海大学 | Method for constructing event map for safe operation of dam |
| CN113535917A (en) * | 2021-06-30 | 2021-10-22 | 山东师范大学 | Intelligent question-answering method and system based on travel knowledge map |
| CN113806513A (en) * | 2021-09-30 | 2021-12-17 | 中国人民解放军国防科技大学 | Question-answering system construction method and system based on knowledge graph in military field |
| CN114416939A (en) * | 2021-12-28 | 2022-04-29 | 广州市瑞行科技有限公司 | Intelligent question and answer method, device, equipment and storage medium |
Non-Patent Citations (4)
| Title |
|---|
| RONGXUAN GAO等: "Knowledge question-answering system based on Knowledge Graph of Traditional Chinese Medicine", 《2020 IEEE 9TH JOINT INTERNATIONAL INFORMATION TECHNOLOGY AND ARTIFICIAL INTELLIGENCE CNFERENCE》 * |
| 王宏宇等: "基于军事领域知识图谱的智能问答***设计与实现", 《装甲兵学报》 * |
| 王震南: "基于知识图谱的军事装备知识问答***研究与实现", 《中国优秀硕士学位论文全文数据库社会科学Ⅰ辑》 * |
| 陈晓军等: "企业风险知识图谱的构建及应用" * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116738572A (en) * | 2023-06-19 | 2023-09-12 | 北京知元创通信息技术有限公司 | Aircraft configuration modularized assembly method and system |
| CN117093679A (en) * | 2023-06-19 | 2023-11-21 | ***科技（杭州）有限公司 | Large language model intelligent inquiry dialogue method, system, equipment and medium |
| CN116738572B (en) * | 2023-06-19 | 2024-03-01 | 北京知元创通信息技术有限公司 | Aircraft configuration modularized assembly method and system |
| CN117093679B (en) * | 2023-06-19 | 2024-04-02 | ***科技（杭州）有限公司 | Large language model intelligent inquiry dialogue method, system, equipment and medium |
| CN116595339A (en) * | 2023-07-19 | 2023-08-15 | 东方空间技术(山东)有限公司 | Intelligent processing method, device and equipment for space data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115599899B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108804521B (en) | Knowledge graph-based question-answering method and agricultural encyclopedia question-answering system | |
| CN117033608B (en) | Knowledge graph generation type question-answering method and system based on large language model | |
| CN106874378B (en) | Method for constructing knowledge graph based on entity extraction and relation mining of rule model | |
| CN112989005B (en) | Knowledge graph common sense question-answering method and system based on staged query | |
| CN115599899B (en) | Intelligent question-answering method, system, equipment and medium based on aircraft knowledge graph | |
| CN111931506B (en) | Entity relationship extraction method based on graph information enhancement | |
| CN110990590A (en) | Dynamic financial knowledge map construction method based on reinforcement learning and transfer learning | |
| CN111444700A (en) | Text similarity measurement method based on semantic document expression | |
| CN112650840A (en) | Intelligent medical question-answering processing method and system based on knowledge graph reasoning | |
| CN111783394A (en) | Training method of event extraction model, event extraction method, system and equipment | |
| CN110765277B (en) | Knowledge-graph-based mobile terminal online equipment fault diagnosis method | |
| CN102663129A (en) | Medical field deep question and answer method and medical retrieval system | |
| CN112328766B (en) | Knowledge graph question-answering method and device based on path search | |
| CN113569023A (en) | Chinese medicine question-answering system and method based on knowledge graph | |
| CN111143574A (en) | Query and visualization system construction method based on minority culture knowledge graph | |
| CN110688489A (en) | Knowledge graph deduction method and device based on interactive attention and storage medium | |
| CN110781681A (en) | Translation model-based elementary mathematic application problem automatic solving method and system | |
| CN114818717A (en) | Chinese named entity recognition method and system fusing vocabulary and syntax information | |
| CN112926325A (en) | Chinese character relation extraction construction method based on BERT neural network | |
| CN113468891A (en) | Text processing method and device | |
| CN111666374A (en) | Method for integrating additional knowledge information into deep language model | |
| CN112579666A (en) | Intelligent question-answering system and method and related equipment | |
| CN117786052A (en) | Intelligent power grid question-answering system based on domain knowledge graph | |
| CN116737911A (en) | Deep learning-based hypertension question-answering method and system | |
| CN113468311B (en) | Knowledge graph-based complex question and answer method, device and storage medium |
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 |