A problem analysis method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a question analysis method and device, electronic equipment and storage medium, relates to the technical field of Internet, and particularly relates to the field of knowledge graph. The specific implementation scheme is: obtaining a question input by a user; performing knowledge classification on the question according to a pre-constructed classification type to obtain classification information of the question; performing knowledge understanding on the question to obtain subject segment understanding information and predicate segment understanding information of the question; and generating final analysis information of the question based on the subject segment understanding information and the predicate segment understanding information and the classification information. The embodiments of the present application can provide a general analysis scheme for knowledge class Query, can be applied in different scenes with knowledge class demand, and have stronger explainability and higher accuracy.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, and further to knowledge graph technology, particularly a problem parsing method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the rapid development of the internet, search engines have become increasingly powerful, and users' expectations of them have also risen, shifting from basic retrieving relevant web pages to intelligent question answering. Users primarily express their search intent through the questions they input (Query), therefore, the accuracy of the search engine's understanding of this intent is crucial.

[0003] In existing technologies, traditional problem-solving methods mainly include the following four types:

[0004] First, information in the query is identified through pattern matching. The query is matched against the pattern and its corresponding dictionary; a successful match is used as the parsed information. However, this approach suffers from generalization problems. When migrating to a new domain or problem type, the pattern needs to be reconstructed, and the pattern itself is unlikely to cover all situations.

[0005] Second, the query graph-based approach. This involves first identifying the main entities in the query, then constructing a query graph using methods such as scoring candidate query graphs, and finally using this graph as the understanding result of the query. This approach is relatively complex, and its accuracy in academia does not meet the requirements for practical application.

[0006] Third, the intent recognition and slot filling solution. This involves pre-defining intent types and their corresponding slots, then first identifying the intent type, and then identifying the slot information required for that intent. This solution often requires design and development for a specific parsing problem and is not a system that can be universally applied after a single development effort.

[0007] Fourth, a text2sql query statement is generated end-to-end using a deep learning model. This method uses a deep learning model, with the query as input and the SQL statement as the model's output. Training is performed using query-SQL pairs as training data in an end-to-end manner. However, this approach has poor interpretability and cannot provide detailed information for downstream tasks.

[0008] It can be seen that the above solutions can only solve specific problems and cannot provide a general parsing solution for knowledge-based queries. When migrating domains or question types, a large amount of repetitive development work is required. Summary of the Invention

[0009] This disclosure provides a problem-solving method, apparatus, electronic device, and storage medium.

[0010] Firstly, this application provides a problem-solving method, the method comprising:

[0011] The problem of obtaining user input;

[0012] The problem is classified according to a pre-built classification type to obtain the classification information of the problem;

[0013] By performing knowledge understanding on the problem, we can obtain the subject fragment understanding information and the predicate fragment understanding information of the problem;

[0014] Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, the final analysis information for the question is generated.

[0015] Secondly, this application provides a problem analysis apparatus, the apparatus comprising: an acquisition module, a knowledge classification module, a knowledge understanding module, and an analysis module; wherein,

[0016] The acquisition module is used to acquire the question input by the user;

[0017] The knowledge classification module is used to classify the problem according to a pre-built classification type to obtain the classification information of the problem;

[0018] The knowledge understanding module is used to perform knowledge understanding on the question and obtain the subject fragment understanding information and predicate fragment understanding information of the question;

[0019] The parsing module is used to generate the final parsing information for the question based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information.

[0020] Thirdly, embodiments of this application provide an electronic device, including:

[0021] One or more processors;

[0022] Memory, used to store one or more programs.

[0023] When the one or more programs are executed by the one or more processors, the one or more processors implement the problem parsing method described in any embodiment of this application.

[0024] Fourthly, embodiments of this application provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the problem parsing method described in any embodiment of this application.

[0025] Fifthly, a computer program product is provided that, when executed by a computer device, implements the problem-solving method described in any embodiment of this application.

[0026] The technology described in this application solves the problem that existing technologies can only solve specific problems and cannot provide a general solution for knowledge-based queries. This also means that a large amount of repetitive development work is required when migrating domains or question types. The technical solution provided in this application can provide a general parsing solution for knowledge-based queries, which can be applied to different scenarios with knowledge-based requirements. It is more interpretable and more accurate.

[0027] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0028] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0029] Figure 1 This is a first flowchart illustrating the problem analysis method provided in the embodiments of this application;

[0030] Figure 2 This is a schematic diagram of the problem-solving system provided in this application;

[0031] Figure 3 This is a second flowchart illustrating the problem analysis method provided in the embodiments of this application;

[0032] Figure 4 This is a third flowchart illustrating the problem analysis method provided in the embodiments of this application;

[0033] Figure 5 This is a schematic diagram of the problem analysis device provided in the embodiments of this application;

[0034] Figure 6 This is a block diagram of an electronic device used to implement the problem-solving method of the embodiments of this application. Detailed Implementation

[0035] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0036] Example 1

[0037] Figure 1 This is a first flowchart illustrating the problem-solving method provided in this application embodiment. This method can be executed by a problem-solving device or electronic device, which can be implemented in software and / or hardware, and can be integrated into any smart device with network communication capabilities. Figure 1 As shown, the problem-solving method may include the following steps:

[0038] S101, Problem of obtaining user input.

[0039] In this step, the electronic device can obtain the question input by the user. Specifically, the electronic device can obtain the question input by the user through a search engine or other means. The question in this embodiment can be a knowledge-based question, which is a question raised within a knowledge context. Here, a knowledge context refers to a scenario where the expression of user needs and goals revolve around knowledge; and knowledge refers to objective knowledge content that a knowledge graph can express and include, including entities, concepts, attributes, relationships, formulas, common sense, and other components. This application can be applied to all scenarios where knowledge-based needs are satisfied. Typical scenarios may include: knowledge-based question-and-answer scenarios, and it is also applicable to knowledge-based text dialogue scenarios, multimodal-to-text knowledge retrieval scenarios, etc.

[0040] Figure 2 This is a schematic diagram of the problem-solving system provided in this application. For example... Figure 2As shown, the system can include three logical layers and five core modules. The three logical layers are: knowledge parsing, knowledge understanding, and structured representation. The five core modules are: a knowledge classification module, a knowledge component annotation module, a component association understanding module, a knowledge component alignment module, and a structured output module. Among these modules, the knowledge classification module and the knowledge component annotation module belong to the knowledge parsing logical layer; the component association understanding module and the knowledge component alignment module belong to the knowledge understanding logical layer; and the structured output generation module belongs to the structured representation logical layer. Specifically, the knowledge classification module is used to perform basic knowledge parsing on the question to obtain predefined classification features. It should be noted that, in addition to the topic domain classification, knowledge type classification, and answer type classification given above, this application embodiment may also include other types of classification. Among the five core modules, the knowledge component annotation module is used to perform knowledge parsing on the question's knowledge components to obtain predefined knowledge components. The component association understanding module is used to further annotate and parse the knowledge component text fragments, parsing the knowledge component fragments into structured topic-based parsing information. The knowledge component alignment module aligns the topic-based information generated by the component association understanding module and assigns IDs to parts that can be assigned. The structured output module integrates the information from all the above modules and generates the final parsed information using certain strategies and rules.

[0041] Specifically, the ID-based processing in this application refers to selecting the concept that expresses the more precise or more generalized meaning based on parameter configuration when two meanings belong to the same concept but have different levels of precision. For example, assuming meanings A and B belong to the same concept but have different levels of precision, when a more precise concept is required, the ID-based processing selects concept A; when a more generalized concept is required, the ID-based processing selects concept B. For instance, when a more precise concept is required, in the question "What year was the reclamation of Nanniwan?", P represents "What year was the reclamation?", where the question element is "What year?", which, after alignment, becomes "Year", and the subject element is "Time". Both "Year" and "Time" represent time concepts, but "Year" is a more specific time, so the final subject used is "Year", and the final ID-based result is "Reclamation_Year". Conversely, when a more generalized concept is required, the final result is "Reclamation_Time".

[0042] S102. Classify the problem according to the pre-built classification type to obtain the classification information of the problem.

[0043] In this step, the electronic device can classify the question according to pre-built classification types to obtain the question's classification information. Specifically, this application defines the classification types in different dimensions, including: subject area classification, knowledge type classification, and answer type classification. Subject area classification can include types such as medical, entertainment figures, education, and military; knowledge type classification can include 56 types such as SPO question-and-answer type, limited entity type, enumeration element type, calculation type, and "why" type; answer type classification can include types such as person, time, location, and quantity. It should be noted that, in addition to the subject area classification, knowledge type classification, and answer type classification given above, this application embodiment can also include other types of classification. For example, assuming the user inputs the question "When was the world-famous Dujiangyan Irrigation System built?", this question can be classified as "time" in the "question type" dimension; "scenic spot" in the "domain" dimension; and "time" in the "answer type" dimension. In this step, the electronic device can classify the question through the knowledge classification module, which focuses on the definition and technical implementation of knowledge types. This application provides a complete definition of knowledge types for the first time; it also supports the introduction of classification features with scalable dimensions. Specific implementation technologies include, but are not limited to: 1. Deep learning text classification technology based on pre-trained language models and fine-tuning; 2. Text classification technology based on knowledge annotation, for example, domain classification using conceptual information of entities in a knowledge graph; 3. Intent detection technology based on keyword recognition; 4. Intent detection technology based on pattern matching.

[0044] S103. Perform knowledge comprehension on the problem to obtain comprehension information on the subject fragment and the predicate fragment.

[0045] In this step, the electronic device can perform knowledge understanding on the question, obtaining understanding information for the subject and predicate segments. Specifically, the electronic device can first annotate the question with knowledge components to obtain the subject and predicate information; then, it can perform component association understanding on the subject and predicate information respectively to obtain understanding information for the subject and predicate segments.

[0046] S104. Based on the subject fragment comprehension information, predicate fragment comprehension information, and classification information, generate the final analysis information for the question.

[0047] In this step, the electronic device can generate the final parsed information for the question based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information. Specifically, the electronic device can first generate a structured query statement or a semi-structured query statement corresponding to the question based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information; then, it can use the structured query statement or the semi-structured query statement corresponding to the question to query the question and obtain the final parsed information for the question.

[0048] The problem parsing method proposed in this application first obtains the user-inputted question; then, it categorizes the question according to a pre-built classification type to obtain the question's classification information; next, it performs knowledge understanding on the question to obtain the subject fragment understanding information and predicate fragment understanding information; finally, based on the subject fragment understanding information, predicate fragment understanding information, and classification information, it generates the final parsed information for the question. In other words, this application can generate the final parsed information for a user-inputted question by performing knowledge classification and knowledge understanding. In contrast, existing problem parsing methods can only solve specific problems and cannot provide a general parsing solution for knowledge-based queries. Because this application employs a technical approach of knowledge classification and understanding of the problem, generating the final parsing information of the problem based on subject fragment understanding information, predicate fragment understanding information, and classification information, it overcomes the technical problem that existing technologies can only solve specific problems and cannot provide a general solution for knowledge-based queries. Furthermore, it requires significant repetitive development work when migrating domains or problem types. The technical solution provided by this application can provide a general parsing solution for knowledge-based queries, applicable to different scenarios with knowledge-based requirements, offering stronger interpretability and higher accuracy. Moreover, the technical solution of this application's embodiments is simple and convenient to implement, easy to popularize, and has a wider range of applications.

[0049] Example 2

[0050] Figure 3 This is a second flowchart illustrating the problem analysis method provided in this application embodiment. Further optimizations and extensions can be made based on the above technical solution, and it can be combined with the various optional implementation methods described above.

[0051] like Figure 3 As shown, the problem-solving method may include the following steps:

[0052] S301, Problem of obtaining user input.

[0053] S302. Classify the problem according to the pre-built classification type to obtain the classification information of the problem.

[0054] S303. Mark the knowledge components of the question to obtain the subject and predicate information of the question.

[0055] In this step, the electronic device can annotate the question with knowledge components to obtain the subject and predicate information. Specifically, the electronic device can annotate the question with knowledge components through the knowledge component annotation module to obtain predefined knowledge components. This application first defines the knowledge components, including: entity fragments (e.g., "eating apples" in "What are the benefits of eating apples?"), P fragments (e.g., "China's release time" in "The release time of 'The Wrath of Man'"), and constraint text fragments (e.g., "now" in "Now China launches Shenzhou No. X"). This module focuses on the definition and technical implementation of knowledge components; it provides a complete type definition for knowledge components. The fragments defined by this module differ from those defined by traditional methods; the fragments defined by this module refer to the largest fragment of each element at the current top level, rather than fine-grained fragments without hierarchical levels. Through multiple nested parsing calls, complex multi-hop problems are gradually parsed and understood. This module's output, which parses knowledge components only at the current top level, not only enables recursive parsing of multi-hop problems but also generates multi-level nested patterns. It solves the problem of traditional parsing methods flattening elements, resulting in a single-layer pattern and an exponential increase in the number of patterns. For example, to match an N-hop pattern, traditional methods require configuring a pattern for each number of hops. If some elements also carry constraint information, the number of patterns increases exponentially. However, this module avoids this problem through nested patterns. Furthermore, combined with the parsing results, a pattern-corresponding dictionary can be mined. The module's output can be easily and accurately implemented using a nested pattern parsing approach. Specific technologies include, but are not limited to: 1. Entity recognition technology based on deep learning text sequence annotation using pre-trained language models and fine-tuning; 2. Dictionary-based matching detection technology; 3. Element detection technology based on pattern matching.

[0056] S304. Perform component association understanding on the subject information and predicate information respectively to obtain subject fragment understanding information and predicate fragment understanding information.

[0057] In this step, the electronic device can perform component association understanding on the subject and predicate information respectively, obtaining subject fragment understanding information and predicate fragment understanding information. For example, suppose the user inputs the question "When was the world-famous Dujiangyan Irrigation System built?" The subject information of this question is "the world-famous Dujiangyan Irrigation System"; the predicate information is "when was it built?" The subject fragment understanding information of this question can include the following two parts: the first part is: Dujiangyan Irrigation System (representing the project and topic); the second part is: civilized world (representing the limiting condition). The predicate fragment understanding information of this question can include the following two parts: the first part is: when (representing the interrogative word and question); the second part is: built (representing the event and the event related to the topic).

[0058] Specifically, electronic devices can perform component association understanding on subject and predicate information separately through a component association understanding module. This module further annotates and parses knowledge component text fragments, converting them into structured topic-based parsing information. First, topic-based parsing information is defined, including: topic elements and their corresponding type information, various restrictive information and their corresponding type information, core events and their corresponding subjects and objects, and question elements and their corresponding type information. The module's key feature is the introduction of the concept of topic-based information for the first time, along with the definition and technical implementation of its components. The module's definition of topic elements can cover various constrained query requirements, enabling this solution to address the understanding of queries with complex constraints. Specific implementation technologies include, but are not limited to: 1. Identifying specific components and relationships between components through knowledge graph concept annotation and dependency analysis; 2. Directly identifying specific components and relationships between components through models; 3. Generating topic-based parsing information using simple rules based on elements and their relationships.

[0059] S305. Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, generate the final analysis information for the question.

[0060] The problem parsing method proposed in this application first obtains the user-inputted question; then, it categorizes the question according to a pre-built classification type to obtain the question's classification information; next, it performs knowledge understanding on the question to obtain the subject fragment understanding information and predicate fragment understanding information; finally, based on the subject fragment understanding information, predicate fragment understanding information, and classification information, it generates the final parsed information for the question. In other words, this application can generate the final parsed information for a user-inputted question by performing knowledge classification and knowledge understanding. In contrast, existing problem parsing methods can only solve specific problems and cannot provide a general parsing solution for knowledge-based queries. Because this application employs a technical approach of knowledge classification and understanding of the problem, generating the final parsing information of the problem based on subject fragment understanding information, predicate fragment understanding information, and classification information, it overcomes the technical problem that existing technologies can only solve specific problems and cannot provide a general solution for knowledge-based queries. Furthermore, it requires significant repetitive development work when migrating domains or problem types. The technical solution provided by this application can provide a general parsing solution for knowledge-based queries, applicable to different scenarios with knowledge-based requirements, offering stronger interpretability and higher accuracy. Moreover, the technical solution of this application's embodiments is simple and convenient to implement, easy to popularize, and has a wider range of applications.

[0061] Example 3

[0062] Figure 4 This is a schematic diagram of the third process of the problem analysis method provided in this application embodiment. Further optimizations and extensions can be made based on the above technical solution, and it can be combined with the various optional implementation methods described above. For example... Figure 4 As shown, the problem-solving method may include the following steps:

[0063] S401, Problem of obtaining user input.

[0064] S402. Classify the problem according to the pre-built classification type to obtain the classification information of the problem.

[0065] S403. Label the knowledge components of the question to obtain the subject and predicate information of the question.

[0066] S404. Perform component association understanding on the subject information and predicate information respectively to obtain subject fragment understanding information and predicate fragment understanding information.

[0067] S405. Align the subject fragment comprehension information and the predicate fragment comprehension information separately to obtain the alignment results of the subject fragment comprehension information and the predicate fragment comprehension information.

[0068] In this step, the electronic device can perform alignment processing on the subject fragment understanding information and the predicate fragment understanding information respectively, obtaining the alignment processing results of the subject fragment understanding information and the predicate fragment understanding information. Specifically, the electronic device can first perform disambiguation processing on the subject fragment understanding information to obtain the disambiguation processing result of the subject fragment understanding information; then, based on the disambiguation processing result of the subject fragment understanding information, it can perform disambiguation processing on the predicate fragment understanding information to obtain the disambiguation processing result of the predicate fragment understanding information; finally, the disambiguation processing results of the subject fragment understanding information and the predicate fragment understanding information are used as the alignment processing results of the subject fragment understanding information and the predicate fragment understanding information, respectively. The disambiguation processing in this embodiment is the elimination of ambiguity processing. Word meaning, sentence meaning, and discourse meaning levels will produce different meanings according to different contexts. Disambiguation refers to the process of determining the semantics of an object based on the context. Word meaning disambiguation is semantic disambiguation at the word level and is often used in search engines, opinion mining, text understanding and generation, and reasoning. For example, there are two main methods for semantic disambiguation. One is a statistical method, which involves training a model to perform semantic disambiguation by labeling training corpora. The other is a semantic disambiguation method based on syntactic analysis, which involves performing syntactic analysis on the input sentence and then disambiguating the semantics based on the syntactic analysis. For example, the question word "how big" disambiguates the S for the person type as "age"; the S for the object type as "size"; and the S for the building type as "area".

[0069] Optionally, the disambiguation strategy in this application defines the following four disambiguation dictionaries and corresponding disambiguation tasks: Question element disambiguation dictionary: defines a set of candidate question types corresponding to type S, a set of candidate question types corresponding to events, and a set of candidate question types corresponding to topics, used to disambiguate question elements based on type S, event, and topic; Event element disambiguation dictionary: defines a set of candidate event types corresponding to type S, used to disambiguate event elements based on type S; Topic element disambiguation dictionary: defines a set of candidate topics corresponding to type S and a set of candidate topics corresponding to events, used to disambiguate topic elements based on type S and event type; Constraint disambiguation dictionary: a candidate constraint dictionary generated based on specific entities, used to disambiguate constraints and correct types based on entity information. Specific technical solutions for disambiguation using dictionaries include, but are not limited to: 1. Direct mapping disambiguation through dictionary matching in simple cases; 2. Semantic matching of elements and candidate options based on a pre-trained model + fine-tuning. Furthermore, the electronic device can perform alignment processing on the subject fragment understanding information and the predicate fragment understanding information respectively through the knowledge component alignment module. This module aligns the topic-based information generated by the component association understanding module and IDs the parts that can be IDted. This application first defines the alignment task, including entity alignment, P-information alignment, and temporal information alignment. Entity alignment refers to linking the topic-based information representing an entity to the encyclopedia entity; P-information alignment refers to normalizing reasoning logic or statements, for example, "release time," "release date," and "when it will be released" are all normalized to "release time"; temporal information alignment refers to parsing temporally related information into a unified expression form, for example, "second son," "second son," and "second son" are all expressed as "son (n=2)." This module focuses on defining the alignment task and its technical implementation. For the first time, P-information alignment is formally defined and technically implemented. Specific technologies for implementing the entity alignment task include, but are not limited to: 1. Encyclopedia card-generating results information based on online Baidu search queries + statistics + simple rules; 2. A matching model between encyclopedia entity descriptions and queries implemented based on pre-trained models and fine-tuning. The specific technologies for aligning time-series information include, but are not limited to: 1. Pattern-based parsing; 2. Dictionary-based parsing.Specific technologies for achieving P information alignment include, but are not limited to: 1. Element normalization strategy: Normalizing various elements in structured information into predefined expressions using a dictionary, for example, "father," "dad," and "old man" are all normalized to "father"; "time," "year and month," "date," and "when" are all normalized to "time"; 2. Information disambiguation strategy: Disambiguating information based on the parsed information of S entities, for example, the question word "how big"; disambiguating S for people into "age"; disambiguating S for objects into "size"; disambiguating S for buildings into "area"; 3. ID transformation strategy: Converting structured information into predefined standard format P information based on the information in the normalized structured information (e.g., including keywords, related events, questions, limiting elements, etc.) as well as S information and Query knowledge type information.

[0070] The information disambiguation strategy in this application focuses on defining and constructing the disambiguation knowledge (disambiguation dictionary) used during disambiguation. The disambiguation dictionary construction process utilizes entity concept information from the knowledge graph and the corresponding knowledge information. The disambiguation strategy defines the following four disambiguation dictionaries and corresponding disambiguation tasks: Question Element Disambiguation Dictionary: Defines a set of candidate question types corresponding to type S, a set of candidate question types corresponding to events, and a set of candidate question types corresponding to topics, used to disambiguate question elements based on type S, event, and topic; Event Element Disambiguation Dictionary: Defines a set of candidate event types corresponding to type S, used to disambiguate event elements based on type S; Topic Element Disambiguation Dictionary: Defines a set of candidate topics corresponding to type S and a set of candidate topics corresponding to events, used to disambiguate topic elements based on type S and event type; Constraint Disambiguation Dictionary: A candidate constraint dictionary generated based on specific entities, used to disambiguate constraints and correct types based on entity information. Specific technical solutions for disambiguation using dictionaries include, but are not limited to: 1. Direct mapping disambiguation through dictionary matching in simple cases; 2. Semantic matching techniques between elements and candidate options implemented based on pre-trained models and fine-tuning.

[0071] The IDization strategy in this application focuses on the definition of the IDization concept and the definition and implementation of some special processes during the IDization process. Key core process definitions include: a PK strategy between the question element and the original topic element: when two meanings belong to the same concept but have different levels of precision, the more precise or more generalized meaning is selected based on parameter configuration. For example, when a more precise concept is required, in "When was the Nanniwan reclamation time?", P is "When was the reclamation time?", where the question element is "which year", which, after alignment, becomes "year", and the topic element is "time". Both "year" and "time" represent time concepts, but "year" is a more specific time, so the final topic used is "year", and the final IDization result is "Reclamation_Year". Conversely, when a more generalized concept is required, the final result is "Reclamation_Time". Element completion strategy: for semantically incomplete cases, missing elements are completed. For example, in "Wrath of Man 2 was released", P is "released on", and the corresponding event is "released", but related topic information is missing. The corresponding topic information "time" is completed according to the configuration. The ID-based approach for P aligns complex and diverse P fragments into a standard format. For P fragments with the same semantics, they will eventually be aligned into a unique string ID. This ID can be used to generate subsequent structured query statements, or the output can be used to analyze the distribution of P requirements in a large number of queries.

[0072] S406. Based on the alignment processing results of the subject fragment comprehension information, the subject fragment comprehension information is corrected to obtain the corrected result of the subject fragment comprehension information.

[0073] S407. Based on the alignment processing results of the predicate fragment comprehension information, the predicate fragment comprehension information is corrected to obtain the corrected result of the predicate fragment comprehension information.

[0074] S408. The correction results of the subject fragment comprehension information and the correction results of the predicate fragment comprehension information shall be used as the subject fragment comprehension information and the predicate fragment comprehension information, respectively.

[0075] S409. Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, generate the final analysis information for the question.

[0076] In this step, the electronic device can generate the final parsed information of the question based on the subject fragment understanding information, predicate fragment understanding information, and classification information. Specifically, the electronic device can generate the final parsed information of the question through the structured output module. This module integrates the information from all the above modules and uses certain strategy rules to generate the final parsed information. This application first defines structured information, including: knowledge parsing information of each dimension of the query, parsing information of each element of the query and aligned information, query statements used to retrieve answers, and a sequence of semi-structured query statements that can be used to retrieve answers. Among them, query statements refer to query statements that can be directly executed on a specific database, such as SQL statements, Gremlin statements, SPARQL, etc.; semi-structured query statements refer to an abstract query statement that cannot be executed on a specific database itself, but defines a specification for a specific query operation, such as 1) finding entities: the find_entity operation, which can carry name information and other constraint information; 2) getting entity attribute information: the get_entity_info operation, which can carry various constraints of this attribute information, such as related events, time limits, space limits, question types, etc. This module focuses on defining structured output information, defining semi-structured query statement sequences, and implementing the technical aspects of parsing information into query statements and semi-structured query statement sequences. Specific technologies include, but are not limited to: 1. If the core elements in the query can be ID-ized, generating structured query statements based on the query classification information and the ID-ized information; 2. Understanding the topic and constraint information of the results based on the fine-grained annotations of the core elements, and generating corresponding semi-structured query information based on the query classification information.

[0077] This application presents a general knowledge-based query understanding solution that can be applied to various fields after implementation. The parsed information obtained by this application has multi-layered understanding and can be applied to various downstream tasks, supporting both structured and semi-structured queries. This application inherently possesses the ability to solve multi-hop problems. This application can solve complex constrained problems. This application can also be used to mine user needs after parsing a large number of queries. This application can be used to generate online parsing patterns and dictionaries for mining various elements in the patterns.

[0078] The problem parsing method proposed in this application first obtains the user-inputted question; then, it categorizes the question according to a pre-built classification type to obtain the question's classification information; next, it performs knowledge understanding on the question to obtain the subject fragment understanding information and predicate fragment understanding information; finally, based on the subject fragment understanding information, predicate fragment understanding information, and classification information, it generates the final parsed information for the question. In other words, this application can generate the final parsed information for a user-inputted question by performing knowledge classification and knowledge understanding. In contrast, existing problem parsing methods can only solve specific problems and cannot provide a general parsing solution for knowledge-based queries. Because this application employs a technical approach of knowledge classification and understanding of the problem, generating the final parsing information of the problem based on subject fragment understanding information, predicate fragment understanding information, and classification information, it overcomes the technical problem that existing technologies can only solve specific problems and cannot provide a general solution for knowledge-based queries. Furthermore, it requires significant repetitive development work when migrating domains or problem types. The technical solution provided by this application can provide a general parsing solution for knowledge-based queries, applicable to different scenarios with knowledge-based requirements, offering stronger interpretability and higher accuracy. Moreover, the technical solution of this application's embodiments is simple and convenient to implement, easy to popularize, and has a wider range of applications.

[0079] Example 4

[0080] Figure 5 This is a schematic diagram of the problem analysis device provided in an embodiment of this application. For example... Figure 5 As shown, the device 500 includes: an acquisition module 501, a knowledge classification module 502, a knowledge understanding module 503, and a parsing module 504; wherein,

[0081] The acquisition module 501 is used to acquire the question input by the user;

[0082] The knowledge classification module 502 is used to classify the problem according to a pre-built classification type to obtain the classification information of the problem;

[0083] The knowledge understanding module 503 is used to perform knowledge understanding on the question and obtain the subject fragment understanding information and predicate fragment understanding information of the question.

[0084] The parsing module 504 is used to generate the final parsing information for the question based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information.

[0085] Furthermore, the knowledge understanding module 503 is specifically used to annotate the knowledge components of the question to obtain the subject information and predicate information of the question; and to perform component association understanding on the subject information and the predicate information respectively to obtain the subject fragment understanding information and the predicate fragment understanding information.

[0086] Furthermore, the device further includes: an alignment module 505 (not shown in the figure), used to perform alignment processing on the subject fragment understanding information and the predicate fragment understanding information respectively, to obtain the alignment processing result of the subject fragment understanding information and the alignment processing result of the predicate fragment understanding information; to correct the subject fragment understanding information based on the alignment processing result of the subject fragment understanding information, to obtain the corrected result of the subject fragment understanding information; to correct the predicate fragment understanding information based on the alignment processing result of the predicate fragment understanding information, to obtain the corrected result of the predicate fragment understanding information; and to use the corrected result of the subject fragment understanding information and the corrected result of the predicate fragment understanding information as the subject fragment understanding information and the predicate fragment understanding information respectively.

[0087] Furthermore, the alignment module 505 is specifically used to disambiguate the subject fragment understanding information to obtain the disambiguation result of the subject fragment understanding information; to disambiguate the predicate fragment understanding information based on the disambiguation result of the subject fragment understanding information to obtain the disambiguation result of the predicate fragment understanding information; and to use the disambiguation result of the subject fragment understanding information and the disambiguation result of the predicate fragment understanding information as the alignment result of the subject fragment understanding information and the alignment result of the predicate fragment understanding information, respectively.

[0088] Furthermore, the alignment module 505 is also used to perform ID-based processing on the disambiguation processing result of the subject fragment understanding information and the disambiguation processing result of the predicate fragment understanding information, respectively, to obtain the ID-based processing result of the subject fragment understanding information and the ID-based processing result of the predicate fragment understanding information; and to use the ID-based processing result of the subject fragment understanding information and the ID-based processing result of the predicate fragment understanding information as the alignment processing result of the subject fragment understanding information and the alignment processing result of the predicate fragment understanding information, respectively.

[0089] Furthermore, the parsing module 504 is specifically used to generate a structured query statement or a semi-structured query statement corresponding to the question based on the subject fragment understanding information, the predicate fragment understanding information, and the classification information; and to use the structured query statement or the semi-structured query statement corresponding to the question to query the question and obtain the final parsing information of the question.

[0090] The aforementioned problem-solving apparatus can execute the methods provided in any embodiment of this application, and possesses the corresponding functional modules and beneficial effects for executing the methods. Technical details not described in detail in this embodiment can be found in the problem-solving methods provided in any embodiment of this application.

[0091] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0092] Example 5

[0093] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0094] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0095] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0096] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0097] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as problem-solving methods. For example, in some embodiments, the problem-solving method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the problem-solving method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the problem-solving method by any other suitable means (e.g., by means of firmware).

[0098] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0099] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0100] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0101] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0102] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0103] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0104] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0105] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A problem analysis method, the method comprising: The problem of obtaining user input; The problem is classified according to a pre-built classification type to obtain the classification information of the problem; The problem is subjected to knowledge understanding to obtain subject fragment understanding information and predicate fragment understanding information; wherein, the subject fragment understanding information and predicate fragment understanding information are obtained by performing component association understanding on the subject information and predicate information respectively to obtain topic-type structured information; the topic-type structured information includes: topic elements and corresponding type information, various restrictive information and corresponding type information, core events and their corresponding subject and object, and question elements and their corresponding type information; Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, the final analysis information for the question is generated.

2. The method of claim 1, wherein, By performing knowledge understanding on the problem, we obtain the subject fragment understanding information and the predicate fragment understanding information of the problem, including: The problem is labeled with knowledge components to obtain the subject and predicate information of the problem; The subject information and the predicate information are respectively subjected to component association understanding to obtain the subject fragment understanding information and the predicate fragment understanding information.

3. The method according to claim 2, further comprising: Alignment processing is performed on the subject fragment comprehension information and the predicate fragment comprehension information respectively to obtain the alignment processing result of the subject fragment comprehension information and the alignment processing result of the predicate fragment comprehension information; The subject fragment comprehension information is corrected based on the alignment processing result of the subject fragment comprehension information to obtain the corrected result of the subject fragment comprehension information; Based on the alignment processing result of the predicate fragment understanding information, the predicate fragment understanding information is corrected to obtain the corrected result of the predicate fragment understanding information; The correction results of the subject fragment comprehension information and the correction results of the predicate fragment comprehension information are respectively used as the subject fragment comprehension information and the predicate fragment comprehension information.

4. The method of claim 3, wherein, Alignment processing is performed on the subject fragment comprehension information and the predicate fragment comprehension information respectively to obtain the alignment processing results of the subject fragment comprehension information and the predicate fragment comprehension information, including: The subject fragment comprehension information is disambiguated to obtain the disambiguation result of the subject fragment comprehension information; Based on the disambiguation processing result of the subject fragment comprehension information, the predicate fragment comprehension information is disambiguated to obtain the disambiguation processing result of the predicate fragment comprehension information; The disambiguation processing results of the subject fragment comprehension information and the disambiguation processing results of the predicate fragment comprehension information are respectively used as the alignment processing results of the subject fragment comprehension information and the predicate fragment comprehension information.

5. The method according to claim 4, further comprising: The disambiguation processing results of the subject fragment comprehension information and the disambiguation processing results of the predicate fragment comprehension information are respectively processed into IDs to obtain the ID processing results of the subject fragment comprehension information and the ID processing results of the predicate fragment comprehension information; The ID-based processing results of the subject fragment comprehension information and the ID-based processing results of the predicate fragment comprehension information are respectively used as the alignment processing results of the subject fragment comprehension information and the predicate fragment comprehension information.

6. The method of claim 1, wherein, Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, the final parsing information for the question is generated, including: Based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information, a structured query statement or a semi-structured query statement corresponding to the question is generated; Use the structured or semi-structured query statement corresponding to the question to query the question and obtain the final parsed information of the question.

7. A problem-solving apparatus, the apparatus comprising: The module comprises an acquisition module, a knowledge classification module, a knowledge understanding module, and a parsing module; among which, The acquisition module is used to acquire the question input by the user; The knowledge classification module is used to classify the problem according to a pre-built classification type to obtain the classification information of the problem; The knowledge understanding module is used to perform knowledge understanding on the question, obtaining subject fragment understanding information and predicate fragment understanding information of the question; wherein, the subject fragment understanding information and predicate fragment understanding information are obtained by performing component association understanding on the subject information and predicate information respectively to obtain topic-type structured information; the topic-type structured information includes: topic elements and corresponding type information, various restrictive information and corresponding type information, core events and the corresponding subject and object of the events, and question elements and corresponding type information; The parsing module is used to generate the final parsing information for the question based on the subject fragment comprehension information, the predicate fragment comprehension information, and the classification information.

8. The apparatus according to claim 7, wherein the knowledge understanding module is specifically configured to annotate the knowledge components of the problem to obtain the subject information and predicate information of the problem; and to perform component association understanding on the subject information and the predicate information respectively to obtain the subject fragment understanding information and the predicate fragment understanding information.

9. The apparatus according to claim 8, further comprising: An alignment module is used to perform alignment processing on the subject fragment comprehension information and the predicate fragment comprehension information respectively, to obtain the alignment processing result of the subject fragment comprehension information and the alignment processing result of the predicate fragment comprehension information; The subject fragment comprehension information is corrected based on the alignment processing result of the subject fragment comprehension information to obtain the corrected result of the subject fragment comprehension information; Based on the alignment processing result of the predicate fragment understanding information, the predicate fragment understanding information is corrected to obtain the corrected result of the predicate fragment understanding information; The correction results of the subject fragment comprehension information and the correction results of the predicate fragment comprehension information are respectively used as the subject fragment comprehension information and the predicate fragment comprehension information.

10. The apparatus according to claim 9, wherein the alignment module is specifically configured to perform disambiguation processing on the subject fragment comprehension information to obtain a disambiguation processing result of the subject fragment comprehension information; perform disambiguation processing on the predicate fragment comprehension information based on the disambiguation processing result of the subject fragment comprehension information to obtain a disambiguation processing result of the predicate fragment comprehension information; and use the disambiguation processing result of the subject fragment comprehension information and the disambiguation processing result of the predicate fragment comprehension information as the alignment processing result of the subject fragment comprehension information and the alignment processing result of the predicate fragment comprehension information, respectively.

11. The apparatus according to claim 10, wherein the alignment module is further configured to perform ID-based processing on the disambiguation processing result of the subject fragment understanding information and the disambiguation processing result of the predicate fragment understanding information respectively, to obtain the ID-based processing result of the subject fragment understanding information and the ID-based processing result of the predicate fragment understanding information; and to use the ID-based processing result of the subject fragment understanding information and the ID-based processing result of the predicate fragment understanding information as the alignment processing result of the subject fragment understanding information and the alignment processing result of the predicate fragment understanding information respectively.

12. The apparatus according to claim 7, wherein the parsing module is specifically configured to generate a structured query statement or a semi-structured query statement corresponding to the question based on the subject fragment understanding information, the predicate fragment understanding information, and the classification information; and to query the question using the structured query statement or the semi-structured query statement corresponding to the question to obtain the final parsing information of the question.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

15. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images