Intelligent question and answer technology standardization system architecture based on knowledge graph

Abstract

The invention discloses an intelligent question and answer technology standardization system architecture based on a knowledge graph, and relates to the technical field of intelligent question and answer. According to the method, a unified knowledge representation standard is established, so that knowledge based on the knowledge graph can be shared and fused among different systems; the processing flow of the intelligent question-answering system is standardized, and the question-answering accuracy and quality are improved; meanwhile, interoperability between different intelligent question-answering systems is enhanced, and wide application of the intelligent question-answering technology is promoted.

Description

Technical Field

[0001] The present invention relates to the field of intelligent question answering technology, and specifically to a standardized system architecture of intelligent question answering technology based on knowledge graph, which aims to improve the accuracy, versatility and interoperability of intelligent question answering systems. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.

[0003] With the rapid development of information technology, intelligent question-answering systems have gone through several stages of development, and knowledge graph intelligent question-answering technology has gradually emerged and developed in this process.

[0004] Early intelligent question-answering systems were mainly based on simple keyword matching and rule bases. This approach can play a certain role in dealing with simple and clear questions, but its ability to handle complex semantics and fuzzy questions is extremely limited because it cannot understand the deep meaning and knowledge association behind the questions.

[0005] Subsequently, machine learning technology began to be applied to the field of intelligent question answering, and the accuracy of answers was improved by training models with large amounts of text data. However, this method still has shortcomings in terms of knowledge representation. The model is difficult to effectively utilize structured knowledge, and the answering effect is poor for questions that require reasoning and integration of multiple aspects of knowledge.

[0006] In this context, knowledge graph technology came into being. Knowledge graphs bring new ideas to intelligent question answering with their structured and semantic knowledge representation. Initially, the construction of knowledge graphs mainly relied on manual or semi-automatic methods. Experts manually extracted and organized knowledge and built small knowledge graphs for question answering systems in specific fields. These systems have shown high accuracy in specific fields and can accurately answer some factual questions using entities and relationships in knowledge graphs.

[0007] With the further development of technology, automated knowledge graph construction methods have gradually matured, including the use of natural language processing technology to extract entities and relationships, and the use of machine learning algorithms to automatically expand knowledge graphs. This has led to the continuous expansion of the scale of knowledge graphs and the coverage of increasingly wider fields. At the same time, the application of knowledge graphs in intelligent question-answering has also evolved from simple factual queries to complex reasoning questions and answers. For example, it is possible to answer questions involving multiple entities and multi-layer relationships, and to answer questions by mining hidden knowledge paths through the reasoning mechanism in the knowledge graph.

[0008] However, despite significant progress in knowledge graph intelligent question answering technology, there are still many problems.

[0009] 1. Lack of unified standards for knowledge representation

[0010] Different intelligent question-answering systems use different ways to represent knowledge. Some use simple text indexes, while others try to build knowledge graphs. However, the construction and representation methods of knowledge graphs are also different, making it difficult to share and integrate knowledge.

[0011] (II) The accuracy and quality of questions and answers vary

[0012] Due to the lack of unified technical standards, it is difficult to ensure the quality of the system in the process of understanding user questions, querying knowledge and generating answers. Some systems may not be able to accurately understand complex questions, or the answers they return may not be accurate or comprehensive enough.

[0013] 3. Poor interoperability

[0014] It is difficult for intelligent question-answering systems developed by different developers to interact and collaborate effectively, which limits the development of intelligent question-answering technology in large-scale, cross-system application scenarios. For example, question-answering systems of different departments within an enterprise cannot be integrated, or question-answering systems in different industries cannot achieve data and function interaction. Summary of the invention

[0015] The purpose of the present invention is to provide a standardized system architecture of intelligent question answering technology based on knowledge graph to solve the problems existing in the prior art and achieve the following goals:

[0016] Establish a unified knowledge representation standard to enable knowledge based on knowledge graphs to be shared and integrated among different systems;

[0017] Standardize the processing flow of the intelligent question-answering system to improve the accuracy and quality of questions and answers;

[0018] Enhance the interoperability between different intelligent question-answering systems and promote the widespread application of intelligent question-answering technology.

[0019] The technical solution of the present invention is as follows:

[0020] A standardized system architecture for intelligent question-answering technology based on knowledge graph, including:

[0021] Knowledge graph construction standard module, question understanding and representation standard module, question-answering reasoning and query standard module, answer generation and evaluation standard module, system interface and interoperability standard module;

[0022] The knowledge graph construction standard module is used to standardize data collection, define entity and relationship extraction standards, and define knowledge graph storage standards;

[0023] The problem understanding and representation standard module is used to standardize the problem preprocessing method and define the problem semantic representation standard;

[0024] The question-answering reasoning and query standard module is used to formulate reasoning rules and define query optimization standards;

[0025] The answer generation and evaluation criteria module is used to standardize answer generation and establish answer evaluation indicators;

[0026] The system interface and interoperability standard module is used to standardize interface design and formulate interoperability protocols.

[0027] Furthermore, standardize data collection, including: prescribing methods and formats for collecting data from different data sources;

[0028] Define the entity and relationship extraction standards, including: define the methods for extracting entities and relationships from the collected data; clarify the types of entities and relationships, and specify the parameters and performance requirements of the extraction algorithms or tools used;

[0029] Define the knowledge graph storage standards, including: determine the storage structure of the knowledge graph and the database selection.

[0030] Furthermore, the question preprocessing method is defined, including: defining standard methods for cleaning, word segmentation, and part-of-speech tagging of user input questions; specifying the natural language processing tools and libraries used and their configuration parameters;

[0031] Define the semantic representation standard for the problem, including: define how to transform the preprocessed problem into a semantic representation that can be understood by a computer, and specify the format and specifications of the semantic representation.

[0032] Furthermore, the method of converting the preprocessed problem into a semantic representation form understandable by a computer includes: based on a vector space model and semantic role labeling.

[0033] Further, formulating reasoning rules includes: formulating unified reasoning rules based on the knowledge and semantic relations in the knowledge graph;

[0034] Define query optimization criteria, including: specify optimization strategies for querying answers in the knowledge graph.

[0035] Furthermore, the inference rules include: transitive inference rules, inherited inference rules, and causal relationship inference rules;

[0036] The optimization strategy includes: a method for constructing query statements, a method for utilizing indexes, and a query process for different types of questions.

[0037] Further, standardizing answer generation includes: determining how to generate accurate and natural answers based on query results;

[0038] Establish answer evaluation indicators, including: establish a set of indicators to evaluate the quality of answers, and specify how to collect user feedback to further improve the answer generation process.

[0039] Furthermore, the format of the answer, the norms of language expression, and the generation methods for different types of answers are determined to generate accurate and natural answers;

[0040] The indicator system includes: accuracy, completeness, relevance and conciseness of the answers.

[0041] Furthermore, the interface design is standardized, including: defining the external interface of the intelligent question-answering system and specifying the definition and usage of interface parameters;

[0042] Develop interoperability protocols, including: developing protocols for interaction between different intelligent question-answering systems.

[0043] Further, the external interface includes: protocols and data formats of input interfaces and output interfaces;

[0044] The protocol for interaction between the different intelligent question-answering systems includes: identity authentication, data exchange format and communication mechanism between systems.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. Improve the efficiency of knowledge sharing and integration

[0047] In the prior art, there is a lack of unified standards for knowledge representation, and knowledge between different systems is difficult to integrate. The present invention establishes a knowledge graph construction standard, with clear specifications from data collection, entity and relationship extraction to storage, so that knowledge from different sources and different construction methods can be integrated into the knowledge graph according to unified standards. This is conducive to the sharing and integration of knowledge, avoiding the problem of knowledge islands, and enterprises or organizations can make more full use of existing knowledge resources and enhance the overall value of knowledge utilization.

[0048] 2. Improve the accuracy and quality of Q&A

[0049] Since the current intelligent question-answering system lacks a unified processing flow standard, the accuracy and quality of questions and answers vary. The question understanding and representation standards, question-answering reasoning and query standards, and answer generation and evaluation standards of the present invention provide specifications for the entire process of the intelligent question-answering system. Through unified question preprocessing, semantic representation methods, reasoning rules, and query optimization strategies, the system can understand user questions more accurately, more effectively use the knowledge in the knowledge graph for reasoning and query, and generate high-quality answers. For example, in the processing of complex problems, reasoning rules can help tap into deep knowledge relationships, thereby giving more comprehensive and accurate answers, effectively reducing incorrect or incomplete answers, and improving user satisfaction.

[0050] 3. Enhance system interoperability

[0051] At present, the interoperability between intelligent question-and-answer systems developed by different developers is poor, which limits the large-scale application of the technology. The system interface and interoperability standards of the present invention design a unified interface and interaction protocol. This allows different intelligent question-and-answer systems to be easily connected and interacted to achieve knowledge sharing and collaborative question-and-answer. Whether it is the system integration of different departments within an enterprise or the system collaboration between different industries, it can be carried out more smoothly. For example, in the cooperation scenario of the medical and insurance industries, the intelligent question-and-answer systems of both parties can share relevant knowledge by following the interoperability protocol to provide users with more comprehensive services, expand the application scope and scenarios of intelligent question-and-answer technology, and promote cross-domain innovative applications.

[0052] 4. Reduce development and maintenance costs

[0053] In the absence of unified standards, when developing a new intelligent question-and-answer system or improving an existing system, developers need to spend a lot of time and energy to deal with problems such as inconsistent knowledge representation and incompatible interfaces. The standardized system of the present invention provides developers with clear guidance and specifications, reducing uncertainty and repetitive work in the development process. At the same time, in terms of system maintenance, the standardized system facilitates unified management and updating, reducing maintenance costs. For example, when it is necessary to update the knowledge in the knowledge graph or improve the question-and-answer algorithm, based on standardized processes and interfaces, modifications and deployments can be made more efficiently, improving the maintainability and scalability of the system.

[0054] 5. Promote technological innovation and development

[0055] The standardization system provides a common framework for the field of intelligent question-answering technology, allowing researchers and developers to focus more on innovation on this basis. They can develop new algorithms, models, and applications based on unified standards without worrying about compatibility issues with other systems. At the same time, the existence of standards is conducive to the dissemination and exchange of knowledge. Different teams can compare and improve technologies on a standardized platform, thereby promoting the rapid development of the entire field of intelligent question-answering technology and accelerating the promotion and application of new technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a standardized system architecture diagram for intelligent question-answering technology based on knowledge graph. DETAILED DESCRIPTION

[0057] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0058] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0059] Embodiment 1

[0060] See also Figure 1 , a standardized system architecture of intelligent question-answering technology based on knowledge graph, specifically including:

[0061] Knowledge graph construction standard module, question understanding and representation standard module, question-answering reasoning and query standard module, answer generation and evaluation standard module, system interface and interoperability standard module;

[0062] The knowledge graph construction standard module is used to standardize data collection, define entity and relationship extraction standards, and define knowledge graph storage standards;

[0063] The problem understanding and representation standard module is used to standardize the problem preprocessing method and define the problem semantic representation standard;

[0064] The question-answering reasoning and query standard module is used to formulate reasoning rules and define query optimization standards;

[0065] The answer generation and evaluation criteria module is used to standardize answer generation and establish answer evaluation indicators;

[0066] The system interface and interoperability standard module is used to standardize interface design and formulate interoperability protocols.

[0067] In this embodiment, specifically, standardizing data collection includes: specifying methods and formats for collecting data from different data sources (such as text files, databases, web pages, etc.), including data format requirements (such as JSON, XML, etc.), data metadata definitions (such as data sources, collection time, etc.), and processing methods for different types of data (such as structured data, semi-structured data, and unstructured data);

[0068] That is, according to the data collection specifications, determine the target data source that needs to collect data. For enterprise-level intelligent question-answering systems, data sources may include the company's internal document management system (such as documents storing corporate rules and regulations, operating manuals, etc.), relational databases (such as employee information databases, project management databases), internal web pages (such as corporate announcements, department introduction pages), etc. Different collection methods are used for different types of data sources. For document management systems, obtain document content through the corresponding API or file reading interface; for databases, use SQL query statements to obtain structured data; for web pages, use web crawler technology to extract text information from web pages;

[0069] During the data collection process, the data is uniformly converted into a specified format, such as JSON format. For each data record, metadata information is added, including the data source (such as the specific document name, database table name, web page URL) and the collection time. For example, an employee record collected from the employee information database: {"name":"Zhang San","department":"R&D Department"}.

[0070] In this embodiment, specifically, defining the entity and relationship extraction standards includes: defining the method of extracting entities and relationships from the collected data; clarifying the types of entities (such as people, organizations, events, etc.) and the types of relationships (such as belonging relationships, causal relationships, etc.), and specifying the parameters and performance requirements of the extraction algorithms or tools used (such as rule-based extraction, machine learning algorithms, etc.);

[0071] That is, extraction is performed based on entity and relationship extraction standards, combined with rule-based and machine learning methods. For enterprise data, first formulate a rule-based extraction template. For example, the relationship between employees and departments is extracted with the text pattern of "name + department". At the same time, a large amount of labeled data is trained using machine learning algorithms to improve the accuracy and generalization of extraction. For the extracted entities, their types are clarified. For example, employees ("Zhang San") belong to person-type entities, and departments ("R&D Department") belong to organization-type entities. Types of relationships include the affiliation relationship between employees and departments, the participation relationship between departments and projects, etc.;

[0072] Check and compare with known standard data to ensure the accuracy of extracted entities and relationships. In the case of incorrect extraction, analyze the cause and adjust the extraction rules or algorithm parameters.

[0073] In this embodiment, specifically, the knowledge graph storage standard is defined, including: determining the storage structure of the knowledge graph and the database selection, for example, a graph database (such as Neo4j) or other databases that support knowledge graph storage can be used, and the storage mode (such as the storage format of nodes and edges, the index establishment method, etc.) is specified to ensure efficient storage and query of the knowledge graph;

[0074] The extracted entities and relationships are stored in a graph database (such as Neo4j) according to the knowledge graph storage standard. In Neo4j, each entity is a node, and the type of entity is the label of the node. For example, the node label of the employee entity "Zhang San" is "Employee", and the node label of the department entity "R&D Department" is "Department". The relationship is an edge, and the type of edge indicates the type of relationship, such as the "BELONGS_TO" edge between an employee and a department;

[0075] Establish appropriate indexes to improve query speed; according to business needs and common query patterns, establish indexes for entity attributes that are often used as query conditions. For example, establish indexes for attributes such as employee names and department names to quickly locate relevant nodes and edges during question-and-answer queries.

[0076] In this embodiment, specifically, the question preprocessing method is defined, including: defining standard methods for cleaning (such as removing noise, special characters, etc.), word segmentation, and part-of-speech tagging of user input questions; specifying the natural language processing tools and libraries (such as NLTK, StanfordNLP, etc.) used and their configuration parameters;

[0077] That is, when the user enters a question, the question is cleaned through the question preprocessing specification. First, remove the extra spaces, punctuation marks (except for punctuation marks that affect semantics, such as question marks) and special characters in the question. For example, "Zhang San, which department is he in?" is processed into "Zhang San is in which department?";

[0078] Then, use the selected natural language processing library (such as NLTK) for word segmentation and part-of-speech tagging. For the question "What projects does Zhang San's department participate in in the company?", the word segmentation result is ["company", "in", "Zhang San", "in", "department", "participated in", "of", "projects", "there are", "which"], and the part-of-speech tagging result is ["NN", "LC", "NN", "VV", "NN", "VV", "DEC", "NN", "VV", "PN"]. In this process, adjust the accuracy of word segmentation and part-of-speech tagging according to the configuration parameters of the natural language processing library, for example, set appropriate vocabulary and language model parameters.

[0079] In this embodiment, specifically, defining the semantic representation standard of the problem includes: defining how to convert the preprocessed problem into a semantic representation form understandable by a computer, which may adopt methods based on vector space model, semantic role labeling, etc., and specifying the format and specification of the semantic representation so as to match it with the knowledge in the knowledge graph;

[0080] That is, according to the semantic representation standard of the problem, the preprocessed problem is converted into a semantic representation. When using the vector space model, first build a vocabulary containing enterprise-related vocabulary (such as employees, departments, projects, business processes, etc.). Then, map the keywords in the question to the vector space. For example, the vector representation of the question "What projects does Zhang San's department in the company participate in?" may be a high-dimensional vector, in which the dimensions related to "Zhang San", "department" and "project" have higher values;

[0081] In addition, the semantic role labeling method can also be combined to analyze the semantic roles in the problem. For example, "Zhang San" is the agent, "the department" is the location role related to the agent, and "the project involved" is the target role. In this way, the semantic structure of the problem can be understood more accurately so as to match it with the knowledge in the knowledge graph. In this process, the parameters of the semantic representation model can be continuously optimized to improve the accuracy of the semantic representation, which can be achieved by training and verifying on the labeled data set.

[0082] In this embodiment, specifically, formulating reasoning rules includes: formulating unified reasoning rules based on the knowledge and semantic relationships in the knowledge graph. These rules are used to derive new knowledge from known knowledge in the question-answering process to answer complex questions. For example, transitive reasoning rules (if A is related to B, and B is related to C, then A and C may have a certain relationship), etc., and clarifying the representation form and application conditions of the rules;

[0083] It should be noted that when answering questions, reasoning is performed according to the rules established by the inference rules. For the above questions, the relationships between employees and departments, and departments and projects in the knowledge graph are used to determine the projects in which Zhang San's department is involved through transitive reasoning. For example, if there are relationships such as "Zhang San BELONGS_TO R&D Department" and "R&D Department PARTICIPATES_IN Project A" and "R&D Department PARTICIPATES_IN Project B" in the knowledge graph, transitive reasoning can be used to deduce that the projects in which Zhang San's department is involved are Project A and Project B;

[0084] For more complex reasoning problems, you may need to apply multiple reasoning rules. For example, in problems involving corporate organizational structure and business processes, you may need to apply both inheritance reasoning rules (such as reasoning about the superior department of a department) and causal reasoning rules (such as reasoning about the results caused by a certain business process). In this case, according to the semantics of the problem and the structure of the knowledge graph, select the appropriate combination of reasoning rules and apply them in a certain order.

[0085] In this embodiment, the query optimization criteria are specifically defined, including: specifying the optimization strategy for querying answers in the knowledge graph, including the query statement construction method, the index utilization method, and the query process for different types of questions (such as simple factual questions, complex reasoning questions, etc.) to improve query efficiency and accuracy;

[0086] That is, according to the query optimization standard, construct a query statement to query the relevant project information in the knowledge graph. According to the semantic representation of the problem and the reasoning results, determine the starting node (such as the "Zhang San" node) and the target node (such as the "project" type node) of the query, as well as the relationship type that needs to be traversed (such as "BELONGS_TO" and "PARTICIPATES_IN"). Use Neo4j's query language (such as Cypher) to construct a query statement, for example: "MATCH(p:Person{name:'Zhang San'})[:BELONGS_TO](d:Department)[:PARTICIPATES_IN](pr:Project)RETURN pr";

[0087] During the query process, make full use of the index to quickly locate related nodes and edges. For example, quickly find the "Zhang San" node through the employee name index, and then find the related department and project nodes along the relationship edge. At the same time, adjust the query strategy according to the storage structure and data distribution of the knowledge graph. If the knowledge graph data volume is large, it may be necessary to use optimization technologies such as paging query and distributed query to improve query efficiency and accuracy.

[0088] In this embodiment, specifically, standardizing answer generation includes: determining how to generate accurate and natural answers based on query results, including answer formats (such as text format, voice format, etc.), language expression specifications (such as grammatical correctness, conformity to user habits, etc.), and methods for generating different types of answers (such as specific values, text descriptions, lists, etc.);

[0089] That is, according to the answer generation specification, the queried project information is generated in a natural and accurate way to answer. If the query result is multiple projects, the answer can be "The projects that Zhang San's department is involved in are Project A, Project B, and Project C." Different generation methods are used for different types of answers. If it is a numerical answer, such as querying the number of employees in a company, the answer can be "The number of employees in the company is 500." If it is a text description answer, such as querying the development history of a company, the answer is to organize the description of related historical events according to certain logic and language expression habits;

[0090] When generating answers, pay attention to the standardization of language expression to ensure that the grammar is correct and in line with user habits. For example, avoid using overly professional or uncommon words (unless the user has specific needs) and use easy-to-understand language to express. At the same time, adjust the level of detail and focus of the answer according to the user's question type and contextual information. If the user simply asks about the project name, the answer only needs to list the project name; if the user further asks for detailed information about the project, the answer needs to include more relevant content, such as the start time and goals of the project.

[0091] In this embodiment, specifically, establishing answer evaluation indicators includes: establishing an indicator system for evaluating answer quality, and specifying how to collect user feedback to further improve the answer generation process;

[0092] That is, based on the answer evaluation indicators, collect user feedback on the answer, such as whether the answer meets the requirements, whether it is clear, whether it is complete, etc. You can set a feedback button or evaluation scale in the user interface to allow users to easily evaluate the answer. For example, users can rate the accuracy of the answer from 1 to 5 points and add text comments;

[0093] Analyze the problems with the answers based on user feedback. If a large number of users report that the answers are inaccurate, you need to check whether the knowledge graph is accurate, whether there are problems with the reasoning and query process, and whether the answer generation algorithm needs to be adjusted. If users report that the answers are not clear enough, you need to optimize the language expression and structure of the answers. Based on the analysis results, improve the answer generation process and improve the quality of the answers. For example, adjust the language template in the answer generation algorithm, add more explanatory information, etc.

[0094] In this embodiment, specifically, standardizing the interface design includes: defining the external interface of the intelligent question-answering system, and specifying the definition and usage of interface parameters;

[0095] That is, design the interface of the intelligent question-answering system, such as using RESTful API. For the input interface, specify the format for accepting user questions, for example, through HTTP POST request, send the user question in JSON format to the specified API endpoint. The JSON data in the request body can be {"question":"What projects does Zhang San's department in the company participate in?"};

[0096] For the output interface, the format of the returned answer is specified as JSON. For example, the returned answer data can be {"answer":"The projects that Zhang San's department is involved in include Project A, Project B, and Project C","confidence":0.9}, where "confidence" indicates the confidence of the answer. In the process of interface implementation, use a suitable Web framework (such as Flask, Django, etc.) to process HTTP requests and responses to ensure the stability and performance of the interface.

[0097] In this embodiment, specifically, formulating an interoperability protocol includes: formulating a protocol for interaction between different intelligent question-answering systems, so that they can share knowledge and collaboratively answer questions. This includes identity authentication, data exchange format, and communication mechanism between systems, so as to realize cross-system intelligent question-answering functions;

[0098] That is, if you need to achieve interoperability with other intelligent question-and-answer systems, configure the systems according to the interoperability protocol. First, perform identity authentication between systems, and you can use standard authentication protocols such as OAuth. For example, when the question-and-answer systems of two enterprise departments interact, they obtain access tokens through the OAuth authorization server to ensure that only authorized systems can communicate;

[0099] Then, use the specified data exchange format and communication mechanism to achieve knowledge sharing and collaborative answering of questions. The data exchange format can be a custom format based on XML or JSON, which is used to transmit questions, answers and related knowledge information between systems. The communication mechanism can use message queues (such as RabbitMQ) or direct network communication (such as the request-response mode under the HTTP protocol). During the interoperability implementation process, sufficient testing is carried out to ensure that different systems can interact and collaborate accurately and stably to avoid data loss or incorrect communication.

[0100] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

[0101] This background section is provided to generally present the context of the invention, and the work of the presently named inventors, the work to the extent described in this background section, and aspects of this section that did not constitute prior art at the time of application are neither explicitly nor implicitly admitted to be prior art to the present invention.

Claims

1. A standardized system architecture for intelligent question answering technology based on knowledge graph, characterized in that: include: Knowledge graph construction standard module, question understanding and representation standard module, question-answering reasoning and query standard module, answer generation and evaluation standard module, system interface and interoperability standard module; The knowledge graph construction standard module is used to standardize data collection, define entity and relationship extraction standards, and define knowledge graph storage standards; The problem understanding and representation standard module is used to standardize the problem preprocessing method and define the problem semantic representation standard; The question-answering reasoning and query standard module is used to formulate reasoning rules and define query optimization standards; The answer generation and evaluation criteria module is used to standardize answer generation and establish answer evaluation indicators; The system interface and interoperability standard module is used to standardize interface design and formulate interoperability protocols.

2. According to claim 1, a standardization system architecture of intelligent question answering technology based on knowledge graph, characterized in that: Standardize data collection, including: specifying methods and formats for collecting data from different data sources; Define the entity and relationship extraction standards, including: define the methods for extracting entities and relationships from the collected data; clarify the types of entities and relationships, and specify the parameters and performance requirements of the extraction algorithms or tools used; Define the knowledge graph storage standards, including: determine the storage structure of the knowledge graph and the database selection.

3. According to the standardization system architecture of intelligent question answering technology based on knowledge graph according to claim 1, it is characterized in that: Define question preprocessing methods, including: define standard methods for cleaning, word segmentation, and part-of-speech tagging of user input questions; specify the natural language processing tools and libraries to be used and their configuration parameters; Define the semantic representation standard for the problem, including: define how to transform the preprocessed problem into a semantic representation that can be understood by a computer, and specify the format and specifications of the semantic representation.

4. According to claim 3, a standardization system architecture of intelligent question answering technology based on knowledge graph is characterized in that: The methods used to transform the preprocessed problems into computer-understandable semantic representations include: vector space model and semantic role labeling.

5. According to claim 1, a standardization system architecture of intelligent question answering technology based on knowledge graph, characterized in that: Formulate inference rules, including: formulating unified inference rules based on the knowledge and semantic relations in the knowledge graph; Define query optimization criteria, including: specify optimization strategies for querying answers in the knowledge graph.

6. According to claim 5, a standardization system architecture of intelligent question-answering technology based on knowledge graph, characterized in that: The inference rules include: transitive inference rules, inherited inference rules, and causal relationship inference rules; The optimization strategy includes: a method for constructing query statements, a method for utilizing indexes, and a query process for different types of questions.

7. The standardization system architecture of intelligent question answering technology based on knowledge graph according to claim 1 is characterized in that: Standardized answer generation, including: determining how to generate accurate and natural answers based on query results; Establish answer evaluation indicators, including: establish a set of indicators to evaluate the quality of answers, and specify how to collect user feedback to further improve the answer generation process.

8. The standardization system architecture of intelligent question-answering technology based on knowledge graph according to claim 7 is characterized in that: Determine the answer format, language expression specifications, and methods for generating different types of answers to generate accurate and natural answers; The indicator system includes: accuracy, completeness, relevance and conciseness of the answers.

9. The standardization system architecture of intelligent question-answering technology based on knowledge graph according to claim 1 is characterized in that: Standardize interface design, including: define the external interface of the intelligent question-answering system, and specify the definition and usage of interface parameters; Develop interoperability protocols, including: developing protocols for interaction between different intelligent question-answering systems.

10. The standardization system architecture of intelligent question-answering technology based on knowledge graph according to claim 9 is characterized in that: The external interface includes: protocols and data formats of input and output interfaces; The protocol for interaction between the different intelligent question-answering systems includes: identity authentication, data exchange format and communication mechanism between systems.

Images