CN115080710A - Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof - Google Patents
Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof Download PDFInfo
- Publication number
- CN115080710A CN115080710A CN202210199670.5A CN202210199670A CN115080710A CN 115080710 A CN115080710 A CN 115080710A CN 202210199670 A CN202210199670 A CN 202210199670A CN 115080710 A CN115080710 A CN 115080710A
- Authority
- CN
- China
- Prior art keywords
- question
- knowledge
- answer
- candidate
- retriever
- 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
-
- 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
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields, which analyzes question sentences of row users by constructing a user question sentence analysis module; constructing an entity linker and a mode linker, optimizing a question analysis result of a user, and generating a candidate question combination characteristic list by arranging and combining map data and map modes; constructing a three-level knowledge retriever, and obtaining candidate answers corresponding to the candidate question sentences from a graph database through the three-level knowledge retriever; and constructing a candidate answer processing module, wherein the candidate answer processing module carries out statistical conversion, sorting and natural language transformation on candidate answers through a statistical intention classification model, an answer sorting model and an answer natural language model to generate a final answer and sends the final answer to the user. According to the invention, the question combination characteristic list is generated by analyzing the user question, and the corresponding candidate answers are retrieved by the three-level knowledge retriever, so that the capability of the intelligent question-answering system to adapt to different fields is improved.
Description
Technical Field
The invention belongs to the field of artificial intelligence, relates to information retrieval, knowledge maps, natural language processing and question-answering technologies, and particularly relates to a construction method of an intelligent question-answering system adaptive to knowledge maps in different fields and the intelligent question-answering system.
Background
Information retrieval is an indispensable way for acquiring information and information, and the information retrieval is from ancient book information retrieval to a search engine and is continuously developed to a current natural language question-answering system. The way of early information retrieval is a text or character string based matching method, including the use of "? Wildcards such as "and". sup. "are used to match strings," like "or" ilike "expressions in database query language SQL, and the like. With the development of the internet, search engines represented by Google and hundredths provide a new information retrieval mode, and information is acquired in the form of keywords.
With the development of information extraction and Knowledge Graph and related technologies, a Knowledge Graph-Based intelligent Question Answering (KBQA) technology can provide more accurate and concise Question answers for users. Current research on KBQA is attached to specific knowledge graphs such as DBPedia, etc. In practical applications, an organization or an enterprise often has a plurality of knowledge maps suitable for respective services, such as a service knowledge base map meeting a customer service scene, a human resource map meeting a human resource related application scene, a customer relationship map for departments such as marketing, marketing or customer service, and the like.
Patent number CN108182262A discloses a method and system for constructing an intelligent question-answering system based on deep learning and knowledge maps, which uses a crawler to obtain an inquiry medical data set of the internet, and performs data preprocessing to obtain a data set with tags; constructing a word segmentation dictionary based on the medical field by combining with the electronic medical record of the hospital, and combining the word segmentation dictionary with the medical dictionary to be used as a systematic word segmentation dictionary; constructing a knowledge graph associated with diseases and symptoms, and aligning disease entities and symptom entities; obtaining a tagged data set according to disease entity alignment; constructing a language model based on deep learning; constructing a knowledge graph-based query optimization algorithm combined with user context information; the method is mainly suitable for the question-answering method of the knowledge graph in the medical diagnosis field, and is not suitable for question-answering of the knowledge graph in different fields.
From the viewpoint of user questions, the questions input by the user often relate to a very wide range, various noises may exist at the same time, the expression modes of the same question are different, spelling errors often occur, and even a piece of randomly input text may be possible. Also, knowledge of the questions may come from multiple different maps, or even multiple maps may be used to answer the questions. At present, the question-answering system meeting the real scene is not realized by a single method, and related research is few, so that the intelligent question-answering technology based on the knowledge graph is difficult to realize large-scale rapid commercial application value.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a method for constructing an intelligent question-answering system adaptive to knowledge maps in different fields.
In order to achieve the purpose of the invention, the technical scheme provided by the invention patent is as follows:
a construction method of an intelligent question-answering system adaptive to knowledge graphs in different fields specifically comprises the following steps:
s1, analyzing the question of the user by utilizing the named entity recognition technology, the ending word segmentation technology, the full text retrieval technology, the semantic matching technology and the synonym strategy;
s2, obtaining map data and map modes in the knowledge map, wherein the map data comprise entity values and attribute values, and generating a candidate question combination feature list by arranging and combining the map data and the map modes based on the user question analysis result obtained in S1;
s3, the three-level knowledge retriever comprises a first retriever, a second retriever and a third retriever, and in the knowledge retrieval stage, based on the question combination feature list generated in S2, candidate answers corresponding to candidate questions are obtained from the graph database through the three-level knowledge retriever;
and S4, the candidate answer processing module carries out statistical conversion, sorting and natural language transformation on the candidate answers generated in the S3 through the statistical intention classification model, the answer sorting model and the answer natural language transformation model, generates the final answer and sends the final answer to the user.
The constructing of the user question analysis module in S1 specifically includes the following steps:
s11, establishing a named entity recognition model by using a named entity recognition technology to perform entity recognition of map data on a question of a user, establishing a training model corpus by using the entity, the sentence where related words of relation and attribute of the labeled knowledge map are located and labeled values thereof, and < Subject, Predict, Object > triplets in the constructed knowledge map, and optimizing the effect of the named entity recognition model by the training corpus, wherein the triplets comprise < entity, entity attribute name, entity attribute value > and < entity, relation, entity >;
s12, importing the graph mode into a word segmentation dictionary by using a bus segmentation technology, and extracting words related to the graph mode from the user question by using the word segmentation dictionary, wherein the graph mode comprises an entity type name, an entity attribute name, a relationship type name and a relationship attribute name;
s13, map data are led into a dictionary of a full-text retrieval engine by using a full-text retrieval technology, words related to the map data and attribute values in an enumerated character string form are extracted from a user question by using an elastic search tool, the full-text retrieval technology is based on candidate entity words which are used for recalling named entity recognition, and entity linkage is carried out by using a semantic matching technology, so that accurate entity words related to the map data are obtained;
s14, constructing a configured entity synonym dictionary and map mode synonym dictionary, expanding the detected entity vocabulary and map mode vocabulary through the entity synonym dictionary and map mode synonym dictionary, and carrying out synonym replacement through a synonym strategy to find the entity vocabulary and map mode vocabulary which really exist in the map data.
The question combination feature list in S2 is to arrange and combine the retrieved map data vocabulary and map mode vocabulary to obtain the combination situation of all questions.
The first-stage knowledge retriever in S3 is a rule engine constructed by creating a retrieval question template, and serves as a first-stage knowledge retriever in the knowledge retrieval stage, and in the knowledge retrieval stage, the question template and the target retrieval sentence for intent determination are rapidly configured, and arbitrary combinations of and/or and conditions between different rules are supported, so that multiplexing of the created question templates is realized.
The second-level knowledge retriever in S3 is a second-level knowledge retriever established based on intent analysis, and in the knowledge retrieval stage, when the first-level knowledge retriever cannot retrieve an effective candidate answer, the second-level knowledge retriever operates to construct an intent classification model, the model is trained based on the collected open-source corpus and new-labeled corpus, and secondary retrieval is performed according to the atlas data and the atlas pattern obtained from the knowledge atlas to obtain the candidate answer.
The third-stage knowledge retriever in S3 is the last-stage bottom-of-pocket retriever in the knowledge retrieval stage, and in the knowledge retrieval stage, when the first-stage knowledge retriever and the second-stage knowledge retriever cannot retrieve an effective candidate answer, the third-stage knowledge retriever operates, and the third-stage knowledge retriever performs semantic matching retrieval directly with the three-tuple data of the knowledge graph by using the question combination feature list detected in the question analysis stage, and finally obtains a candidate answer for the question of the user.
The statistical intention classification model in S4 classifies the generated candidate answers into six categories, i.e., total category, non-category, enumerated category, sequential category, contrast category and other category, and the classification guides the subsequent answer conversion behavior.
The answer ranking model in S4 includes two answer types, namely "sequence class" and "other class", and the answer ranking model uses a Bert-based semantic similarity model and a LambdaRank ranking model to reorder the results, and ranks the candidate answers according to the probability.
The natural language modeling in S4 described above converts the final answer into a natural answer sentence that can be understood by the user by using the general answer template and the customized answer template.
An intelligent question-answering system adaptive to knowledge graphs in different fields is realized based on any one of the construction methods of the intelligent question-answering system adaptive to knowledge graphs in different fields, and comprises the following steps:
firstly, inputting a question sentence of a user by the user;
secondly, the user question analysis module is used for preprocessing question sentences input by the user and analyzing the question sentences;
thirdly, an atlas data linker and an atlas mode linker are used for acquiring atlas data and an atlas mode in the knowledge atlas, and a candidate question and sentence combination feature list is generated by arranging and combining the atlas data and the atlas mode;
fourthly, a third-level knowledge retriever performs knowledge retrieval on the basis of the candidate question and sentence combination feature list generated in the second step to obtain candidate answers corresponding to the candidate question and sentences;
fifthly, the candidate answer processing module is used for processing the candidate answers and classifying the generated candidate answers through a statistical intention classification model, the natural language model converts the final answers into natural answer sentences which can be understood by the user through a general answer template and a customized answer template, the answer ranking model reorders the generated candidate answers through a Bert-based semantic similarity model and a Lambdarank ranking model, and the candidate answers are ranked according to the probability to generate the final answers;
and sixthly, feeding back the final answer generated in the fourth step to the user.
The method of the invention obtains the following positive and beneficial effects through practice:
1. according to the method, the training sample is constructed by utilizing the data in the knowledge graph, so that a good named entity recognition effect is realized, and the named entity recognition accuracy is improved.
2. The invention solves the word segmentation performance problem under large-scale map data by utilizing the word segmentation technology of the elastic search and improves the word segmentation efficiency.
3. The invention provides preparation knowledge for the retrieval answers of the subsequent knowledge retriever by arranging and combining the entity values, the attribute values and the map mode data to ask a sentence combination characteristic list, thereby improving the retrieval accuracy.
4. The invention satisfies the basic question-answer effect under knowledge maps in different fields by constructing the three-level knowledge retriever, supports the convenient and flexible rule configuration optimization effect, is suitable for retrieval of different knowledge maps, and improves the applicability of the knowledge retriever.
Drawings
FIG. 1 is a flow chart of the construction of an intelligent question-answering system in the construction method of an intelligent question-answering system adaptive to knowledge maps in different fields according to the present invention.
Fig. 2 is a flow chart of extracting entity information in a question sentence analyzing module in the intelligent question-answering system construction method adaptive to knowledge maps in different fields according to the present invention.
Detailed Description
In order that the objects, aspects and advantages of the invention will become more apparent, the invention will now be described by way of example only, as illustrated in the accompanying drawings. It is to be understood that such description is merely illustrative and not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.
As shown in fig. 1, a method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields specifically includes the following steps:
s1, analyzing the question of the user by utilizing the named entity recognition technology, the ending word segmentation technology, the full text retrieval technology, the semantic matching technology and the synonym strategy;
s2, obtaining map data and map modes in the knowledge map, wherein the map data comprise entity values and attribute values, and generating a candidate question combination feature list by arranging and combining the map data and the map modes based on the user question analysis result obtained in S1;
s3, the three-level knowledge retriever comprises a first retriever, a second retriever and a third retriever, and in the knowledge retrieval stage, based on the question combination feature list generated in S2, candidate answers corresponding to candidate questions are obtained from the graph database through the three-level knowledge retriever;
and S4, the candidate answer processing module carries out statistical conversion, sorting and natural language transformation on the candidate answers generated in the S3 through the statistical intention classification model, the answer sorting model and the answer natural language transformation model, generates the final answer and sends the final answer to the user.
As shown in fig. 2, the constructing of the user question analysis module in S1 specifically includes the following steps:
s11, establishing a named entity recognition model by using a named entity recognition technology to perform entity recognition of map data on a question of a user, establishing a training model corpus by using the entity, the sentence where related words of relation and attribute of the labeled knowledge map are located and labeled values thereof, and < Subject, Predict, Object > triplets in the constructed knowledge map, and optimizing the effect of the named entity recognition model by the training corpus, wherein the triplets comprise < entity, entity attribute name, entity attribute value > and < entity, relation, entity >; training corpora are constructed based on the knowledge graph, so that the generalization capability of the named entity recognition model is improved, and the entity recognition accuracy is improved;
s12, importing the graph mode into a word cutting dictionary by using a pinch word cutting technology, and extracting words related to the graph mode from the user question by using the word cutting dictionary, wherein the graph mode comprises an entity type name, an entity attribute name, a relationship type name and a relationship attribute name;
s13, map data is imported into a dictionary of a full-text search engine by using a full-text search technology, words related to the map data and attribute values in the form of enumerated character strings are extracted from a question of a user by using an elastic search tool, the full-text search technology is based on candidate entity words used for recalling named entity recognition, and entity linkage is carried out by using a semantic matching technology to obtain accurate entity words related to the map data; the semantic matching technology is the prior art;
s14, constructing a configured entity synonym dictionary and map mode synonym dictionary, expanding the detected entity vocabulary and map mode vocabulary through the entity synonym dictionary and map mode synonym dictionary, and carrying out synonym replacement through a synonym strategy to find the entity vocabulary and map mode vocabulary which really exist in the map data.
The question combination feature list in S2 is to arrange and combine the retrieved map data vocabulary and map mode vocabulary to obtain the combination situation of all questions.
The first-stage knowledge retriever in S3 is a rule engine constructed by creating a retrieval question template, and serves as the first-stage knowledge retriever in the knowledge retrieval stage, and in the knowledge retrieval stage, the question template and the target retrieval sentence for intent determination are rapidly configured, and arbitrary combinations of and/or and conditions between different rules are supported, so that multiplexing of the created question template is realized; the development and use efficiency of the question template is improved.
The second-stage knowledge retriever in S3 is a second-stage knowledge retriever established based on intent analysis, and in the knowledge retrieval stage, when the first-stage knowledge retriever cannot retrieve an effective candidate answer, the second-stage knowledge retriever is operated to construct an intent classification model, the model is trained based on the collected open-source corpora and new-labeled corpora, and secondary retrieval is performed according to the graph data and the graph mode obtained from the knowledge graph to obtain a candidate answer; the method is used for solving the problem of insufficient generalization capability of utilizing the first-level knowledge retriever engine to carry out intention judgment.
The third-stage knowledge retriever in S3 is the last-stage bottom-of-pocket retriever in the knowledge retrieval stage, and in the knowledge retrieval stage, when the first-stage knowledge retriever and the second-stage knowledge retriever cannot retrieve an effective candidate answer, the third-stage knowledge retriever operates, and the third-stage knowledge retriever performs semantic matching retrieval directly with the three-tuple data of the knowledge graph by using the question combination feature list detected in the question analysis stage, and finally obtains a candidate answer for the question of the user.
The statistical intention classification model in S4 classifies the generated candidate answers into six categories, i.e., total category, non-category, enumerated category, sequential category, contrast category and other category, and the classification guides the subsequent answer conversion behavior.
The answer ranking model in S4 includes two answer types, namely "sequence class" and "other class", and the answer ranking model uses a Bert-based semantic similarity model and a LambdaRank ranking model to reorder the results, and ranks the candidate answers according to the probability.
The natural language modeling in S4 described above converts the final answer into a natural answer sentence that can be understood by the user by using the general answer template and the customized answer template.
An intelligent question-answering system adaptive to knowledge graphs in different fields is realized based on any one of the construction methods of the intelligent question-answering system adaptive to knowledge graphs in different fields, and comprises the following steps:
firstly, inputting a question sentence of a user by the user;
secondly, the user question analysis module is used for preprocessing question sentences input by the user and analyzing the question sentences;
thirdly, an atlas data linker and an atlas mode linker are used for acquiring atlas data and an atlas mode in the knowledge atlas, and a candidate question and sentence combination feature list is generated by arranging and combining the atlas data and the atlas mode;
fourthly, a third-level knowledge retriever performs knowledge retrieval on the basis of the candidate question and sentence combination feature list generated in the second step to obtain candidate answers corresponding to the candidate question and sentences;
fifthly, the candidate answer processing module is used for processing the candidate answers and classifying the generated candidate answers through a statistical intention classification model, the natural language model converts the final answers into natural answer sentences which can be understood by the user through a general answer template and a customized answer template, the answer ranking model reorders the generated candidate answers through a Bert-based semantic similarity model and a Lambdarank ranking model, and the candidate answers are ranked according to the probability to generate the final answers;
and sixthly, feeding back the final answer generated in the fourth step to the user.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention and not to limit it; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that; modifications to the specific embodiments of the invention or equivalent substitutions for parts of the technical features may be made; without departing from the spirit of the present invention, it is intended to cover all aspects of the invention as defined by the appended claims.
Claims (10)
1. A method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields is characterized by comprising the following steps:
s1, analyzing the question of the user by utilizing the named entity recognition technology, the ending word segmentation technology, the full text retrieval technology, the semantic matching technology and the synonym strategy;
s2, obtaining map data and map modes in the knowledge map, wherein the map data comprise entity values and attribute values, and generating a candidate question combination feature list by arranging and combining the map data and the map modes based on the user question analysis result obtained in S1;
s3, the three-level knowledge retriever comprises a first retriever, a second retriever and a third retriever, and in the knowledge retrieval stage, based on the question combination characteristic list generated in S2, candidate answers corresponding to the candidate questions are obtained from the database through the three-level knowledge retriever;
and S4, the candidate answer processing module carries out statistical conversion, sorting and natural language transformation on the candidate answers generated in the S3 through the statistical intention classification model, the answer sorting model and the answer natural language transformation model, generates the final answer and sends the final answer to the user.
2. The method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields according to claim 1, wherein the step of constructing the user question analysis module in S1 specifically comprises the following steps:
s11, establishing a named entity recognition model by using a named entity recognition technology to perform entity recognition of map data on a question of a user, establishing a training model corpus by using the entity, the sentence where related words of relation and attribute of the labeled knowledge map are located and labeled values thereof, and < Subject, Predict, Object > triplets in the constructed knowledge map, and optimizing the effect of the named entity recognition model by the training corpus, wherein the triplets comprise < entity, entity attribute name, entity attribute value > and < entity, relation, entity >;
s12, importing the graph mode into a word segmentation dictionary by using a bus segmentation technology, and extracting words related to the graph mode from the user question by using the word segmentation dictionary, wherein the graph mode comprises an entity type name, an entity attribute name, a relationship type name and a relationship attribute name;
s13, map data are led into a dictionary of a full-text retrieval engine by using a full-text retrieval technology, words related to the map data and attribute values in an enumerated character string form are extracted from a user question by using an elastic search tool, the full-text retrieval technology is based on candidate entity words which are used for recalling named entity recognition, and entity linkage is carried out by using a semantic matching technology, so that accurate entity words related to the map data are obtained;
s14, constructing a configured entity synonym dictionary and map mode synonym dictionary, expanding the detected entity vocabulary and map mode vocabulary through the entity synonym dictionary and map mode synonym dictionary, and performing synonym replacement through a synonym strategy to find the entity vocabulary really existing in the map data and the vocabulary really existing in the map mode.
3. The method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields as claimed in claim 1, wherein the question combination feature list in S2 is used for arranging and combining retrieved graph data words and graph mode words to obtain all question combination conditions.
4. The method for constructing an intelligent question-answering system adaptive to different-domain knowledge graphs according to claim 1, wherein a first-level knowledge retriever in S3 is a rule engine constructed by creating a retrieval question template, the first-level knowledge retriever is used as the first-level knowledge retriever in a knowledge retrieval stage, the question template and a target retrieval sentence for intention judgment are rapidly configured in the knowledge retrieval stage, and any combination of and or and conditions among different rules is supported to realize multiplexing of the created question template.
5. The method as claimed in claim 1, wherein the second-stage knowledge retriever in S3 is a second-stage knowledge retriever based on establishment of intent analysis, and in the knowledge retrieval stage, when the first-stage knowledge retriever fails to retrieve an effective candidate answer, the second-stage knowledge retriever operates to construct an intent classification model trained based on the collected open-source corpus and new labeled corpus, and performs secondary retrieval based on the atlas data and the atlas pattern obtained from the knowledge atlas to obtain a candidate answer.
6. The method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields as claimed in claim 1, wherein the third-level knowledge retriever in S3 is the last-level bottom-of-pocket retriever in the knowledge retrieval stage, and in the knowledge retrieval stage, when the first-level knowledge retriever and the second-level knowledge retriever cannot retrieve valid candidate answers, the third-level knowledge retriever operates, and the third-level knowledge retriever utilizes a question combination feature list detected in the question analysis stage to directly perform semantic matching retrieval with three-component data of the knowledge graphs, so as to finally obtain candidate answers for the question of the user.
7. The method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields as claimed in claim 1, wherein the statistical intention classification model in S4 classifies the generated candidate answers into six categories, i.e. total category, non-category, enumerated category, sequential category, contrast category and other category, and the classification guides the subsequent answer conversion behavior.
8. The method for constructing the intelligent question-answering system adaptive to knowledge graphs in different fields as claimed in claim 1, wherein the answer ranking model in S4 includes two answer types of "sequence class" and "other class", and the answer ranking model reorders the results by using a Bert-based semantic similarity model and a LambdaRank ranking model, and ranks the candidate answers according to probability.
9. The method for constructing an intelligent question-answering system adaptive to knowledge graphs in different fields according to claim 1, wherein the natural language modeling in S4 converts a final answer into a natural answer sentence that can be understood by a user by using a general answer template and a customized answer template.
10. An intelligent question-answering system adaptive to knowledge graphs in different fields is characterized by being realized based on the construction method of the intelligent question-answering system adaptive to the knowledge graphs in different fields, which comprises the following steps:
firstly, inputting a question sentence of a user by the user;
secondly, the user question analysis module is used for preprocessing question sentences input by the user and analyzing the question sentences;
thirdly, an atlas data linker and an atlas mode linker are used for acquiring atlas data and an atlas mode in the knowledge atlas, and a candidate question and sentence combination feature list is generated by arranging and combining the atlas data and the atlas mode;
fourthly, a third-level knowledge retriever performs knowledge retrieval through the third-level knowledge retriever based on the candidate question combination feature list generated in the second step to obtain candidate answers corresponding to the candidate questions;
fifthly, the candidate answer processing module is used for processing the candidate answers and classifying the generated candidate answers through a statistical intention classification model, the natural language model converts the final answers into natural answer sentences which can be understood by the user through a general answer template and a customized answer template, the answer ranking model reorders the generated candidate answers through a Bert-based semantic similarity model and a Lambdarank ranking model, and the candidate answers are ranked according to the probability to generate the final answers;
and sixthly, an answer sending module for feeding back the sorted final answers to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210199670.5A CN115080710A (en) | 2022-03-01 | 2022-03-01 | Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210199670.5A CN115080710A (en) | 2022-03-01 | 2022-03-01 | Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115080710A true CN115080710A (en) | 2022-09-20 |
Family
ID=83246143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210199670.5A Pending CN115080710A (en) | 2022-03-01 | 2022-03-01 | Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115080710A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116860909A (en) * | 2023-09-04 | 2023-10-10 | 宁波甬恒瑶瑶智能科技有限公司 | Data storage method, system and storage medium based on biochemical knowledge graph |
| CN117216194A (en) * | 2023-11-08 | 2023-12-12 | 天津恒达文博科技股份有限公司 | Knowledge question-answering method and device, equipment and medium in literature and gambling field |
| CN117891929A (en) * | 2024-03-18 | 2024-04-16 | 南京华飞数据技术有限公司 | Knowledge graph intelligent question-answer information identification method of improved deep learning algorithm |
-
2022
- 2022-03-01 CN CN202210199670.5A patent/CN115080710A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116860909A (en) * | 2023-09-04 | 2023-10-10 | 宁波甬恒瑶瑶智能科技有限公司 | Data storage method, system and storage medium based on biochemical knowledge graph |
| CN116860909B (en) * | 2023-09-04 | 2023-12-22 | 宁波甬恒瑶瑶智能科技有限公司 | Data storage method, system and storage medium based on biochemical knowledge graph |
| CN117216194A (en) * | 2023-11-08 | 2023-12-12 | 天津恒达文博科技股份有限公司 | Knowledge question-answering method and device, equipment and medium in literature and gambling field |
| CN117216194B (en) * | 2023-11-08 | 2024-01-30 | 天津恒达文博科技股份有限公司 | Knowledge question-answering method and device, equipment and medium in literature and gambling field |
| CN117891929A (en) * | 2024-03-18 | 2024-04-16 | 南京华飞数据技术有限公司 | Knowledge graph intelligent question-answer information identification method of improved deep learning algorithm |
| CN117891929B (en) * | 2024-03-18 | 2024-05-17 | 南京华飞数据技术有限公司 | Knowledge graph intelligent question-answer information identification method of improved deep learning algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111475623B (en) | Case Information Semantic Retrieval Method and Device Based on Knowledge Graph | |
| CN110765257B (en) | Intelligent consulting system of law of knowledge map driving type | |
| CN109271505B (en) | Question-answering system implementation method based on question-answer pairs | |
| CN110209787B (en) | Intelligent question-answering method and system based on pet knowledge graph | |
| US9727637B2 (en) | Retrieving text from a corpus of documents in an information handling system | |
| CN115080710A (en) | Intelligent question-answering system adaptive to knowledge graphs in different fields and construction method thereof | |
| CN112163077B (en) | Knowledge graph construction method for field question and answer | |
| CN112667799B (en) | Medical question-answering system construction method based on language model and entity matching | |
| CN112328766B (en) | Knowledge graph question-answering method and device based on path search | |
| CN114036281B (en) | Knowledge graph-based citrus control question-answering module construction method and question-answering system | |
| CN113076411B (en) | Medical query expansion method based on knowledge graph | |
| CN112214335B (en) | Web service discovery method based on knowledge graph and similarity network | |
| CN111858896B (en) | Knowledge base question-answering method based on deep learning | |
| CN116127095A (en) | Question-answering method combining sequence model and knowledge graph | |
| CN115599902B (en) | Oil-gas encyclopedia question-answering method and system based on knowledge graph | |
| CN113569023A (en) | Chinese medicine question-answering system and method based on knowledge graph | |
| CN112328800A (en) | System and method for automatically generating programming specification question answers | |
| CN111125316B (en) | Knowledge base question-answering method integrating multiple loss functions and attention mechanism | |
| CN110851584A (en) | Accurate recommendation system and method for legal provision | |
| Kaddari et al. | Biomedical question answering: A survey of methods and datasets | |
| CN112883172B (en) | Biomedical question-answering method based on dual knowledge selection | |
| Li et al. | Approach of intelligence question-answering system based on physical fitness knowledge graph | |
| CN111008285B (en) | Author disambiguation method based on thesis key attribute network | |
| Çelebi et al. | Automatic question answering for Turkish with pattern parsing | |
| CN113868387A (en) | Word2vec medical similar problem retrieval method based on improved tf-idf weighting |
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 |