User intention understanding method and device based on combination of large model and knowledge graph

Abstract

The invention discloses a user intention understanding method and device based on combination of a large model and a knowledge graph, and relates to the technical field of artificial intelligence, and the method comprises the steps: constructing and storing the knowledge graph; when an updating condition is triggered, dynamically updating the knowledge graph to obtain an updated knowledge graph; performing multi-source data fusion and duplicate removal on the updated knowledge graph to obtain a final knowledge graph; based on the final knowledge graph, performing user intention analysis and deep reasoning on data input by a user to obtain a user demand; the user intention analysis and deep reasoning comprise dominant demand analysis, implicit demand mining and semantic complementation and disambiguation. Through innovative fusion of a large model and a knowledge graph technology, deep understanding of the user intention is realized, shallow semantic limitation and static knowledge bottleneck in the prior art are broken through, and deep semantic analysis and dynamic optimization of the user intention can be realized.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and device for understanding user intent based on a combination of a large model and a knowledge graph. Background Art

[0002] With the rapid development of artificial intelligence technology, user intent recognition technology based on natural language processing (NLP) has been widely used in the fields of intelligent customer service, intelligent assistants, etc. However, existing methods still face many bottlenecks in practical applications:

[0003] 1. Shallow understanding of the problem.

[0004] Current systems are mostly based on keyword matching or simple rule reasoning, which cannot truly capture the user's deep implicit needs. For example, "I'm a little anxious recently, do you have any recommended books?" This type of input cannot fully understand the user's intention through simple keyword recognition.

[0005] 2. Lack of dynamic context analysis.

[0006] Many existing systems lack the ability to analyze users' historical conversations and contexts, and are unable to form a complete understanding of user needs, resulting in information fragmentation in multiple rounds of conversations.

[0007] 3. Static knowledge base is difficult to adapt to changing needs.

[0008] The construction and updating of knowledge bases often need to be done manually, and the knowledge is not timely enough, making it difficult to respond to users' dynamic needs and information in emerging fields in a timely manner.

[0009] 4. The knowledge graph and big model technology are separated.

[0010] The technical separation between knowledge graphs and big models limits the intelligence of the system. Traditional knowledge graphs lack dynamic optimization capabilities, while big models lack structured knowledge support, and the synergy between the two has not yet been fully developed.

[0011] Therefore, how to design a solution that can achieve deep semantic analysis and dynamic optimization of user intentions has become a technical problem that urgently needs to be solved in this field. Summary of the invention

[0012] The purpose of this application is to provide a method and device for understanding user intention based on the combination of a large model and a knowledge graph, which can realize in-depth semantic analysis and dynamic optimization of user intention.

[0013] To achieve the above objectives, this application provides the following solutions:

[0014] In a first aspect, the present application provides a method for understanding user intent based on a combination of a large model and a knowledge graph, and the method for understanding user intent based on a combination of a large model and a knowledge graph includes:

[0015] Construct and store a knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: a public knowledge base and user interaction records.

[0016] When an update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph; the update conditions include: new concept identification, domain demand changes, and knowledge conflict detection.

[0017] The updated knowledge graph is subjected to multi-source data fusion and deduplication to obtain a final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure.

[0018] Based on the final knowledge graph, user intention analysis and deep reasoning are performed on the data input by the user to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

[0019] In a second aspect, the present application provides a user intention understanding system based on a combination of a large model and a knowledge graph, and the user intention understanding system based on a combination of a large model and a knowledge graph includes:

[0020] The knowledge graph construction and storage module is used to construct and store the knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: public knowledge base and user interaction records.

[0021] The knowledge graph update module is used to dynamically update the knowledge graph to obtain an updated knowledge graph when the update conditions are triggered; the update conditions include: new concept identification, domain demand changes and knowledge conflict detection.

[0022] The multi-source data fusion and deduplication module is used to perform multi-source data fusion and deduplication on the updated knowledge graph to obtain the final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure.

[0023] The user intention analysis and deep reasoning module is used to perform user intention analysis and deep reasoning on the data input by the user based on the final knowledge graph to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

[0024] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned methods for understanding user intent based on the combination of a large model and a knowledge graph.

[0025] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for understanding user intent based on the combination of a large model and a knowledge graph.

[0026] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for understanding user intent based on the combination of a large model and a knowledge graph.

[0027] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0028] The present application provides a method and device for understanding user intention based on a combination of a big model and a knowledge graph. First, a knowledge graph is constructed and stored; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a big model; the data sources include: a public knowledge base and user interaction records; secondly, when an update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph; the update condition includes: new concept identification, domain demand changes, and knowledge conflict detection; then, the updated knowledge graph is subjected to multi-source data fusion and deduplication to obtain a final knowledge graph; the multi-source data fusion Deduplication includes: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure; finally, based on the final knowledge graph, the user input data is subjected to user intent analysis and deep reasoning to obtain user needs; the user intent analysis and deep reasoning include: explicit demand analysis, implicit demand mining, and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing. This application achieves a deep understanding of user intentions through the innovative integration of large models and knowledge graph technology, breaking through the shallow semantic limitations and static knowledge bottlenecks of the prior art. This method is suitable for scenarios such as intelligent customer service, conversational AI, and recommendation systems, and can significantly improve user experience, with broad market value and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 This is an application environment diagram of a user intent understanding method based on the combination of a large model and a knowledge graph in one embodiment of the present application.

[0031] Figure 2 A flowchart of a method for understanding user intent based on a combination of a large model and a knowledge graph is provided for one embodiment of the present application.

[0032] Figure 3 A schematic diagram of the overall framework of a method for understanding user intent based on a combination of a large model and a knowledge graph provided in one embodiment of the present application.

[0033] Figure 4A schematic diagram of the functional modules of a user intention understanding system based on a combination of a large model and a knowledge graph provided in one embodiment of the present application;

[0034] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0036] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0037] The user intention understanding method based on the combination of a large model and a knowledge graph provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send multiple data sources to the server 104. After the server 104 receives the multiple data sources, the server 104 constructs and stores knowledge graphs for the multiple data sources; when the update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph; the update condition includes: new concept identification, domain demand changes and knowledge conflict detection; multi-source data fusion and deduplication are performed on the updated knowledge graph to obtain a final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment and dynamic update of the graph structure; based on the final knowledge graph, user intention analysis and deep reasoning are performed on the data input by the user to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing. The server 104 can feed back the obtained user demand to the terminal 102. In addition, in some embodiments, the user intention understanding method based on the combination of the large model and the knowledge graph can also be implemented by the server 104 or the terminal 102 alone, such as the terminal 102 can directly understand the user intention based on the data input by the user, or the server 104 can obtain the user input data from the data storage system and understand the user intention based on the user input data.

[0038] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, and tablet computers. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0039] In an exemplary embodiment, Figure 2 and Figure 3 As shown, a method for understanding user intention based on a large model combined with a knowledge graph is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps S1 to S4. Among them:

[0040] S1: Build and store a knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: a public knowledge base and user interaction records.

[0041] S2: When an update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph; the update conditions include: new concept identification, domain requirement changes, and knowledge conflict detection.

[0042] S3: Perform multi-source data fusion and deduplication on the updated knowledge graph to obtain the final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure.

[0043] S4: Based on the final knowledge graph, perform user intention analysis and deep reasoning on the data input by the user to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

[0044] By implementing the above steps S1 to S4, the present application can improve the ability to accurately identify complex and multi-level user intentions; dynamically build and optimize knowledge graphs, automatically extract information from multiple fields through large models, and enrich user background knowledge; implement deep semantic reasoning based on knowledge graphs, and be able to explore users' explicit needs and implicit intentions, and achieve deep semantic analysis and dynamic optimization of user intentions.

[0045] In an exemplary embodiment, in step S1, it specifically includes:

[0046] S11: Based on multiple data sources, obtain initial knowledge and use large models to extract structured information from natural language to generate entity, relationship and semantic aggregation triples.

[0047] S12: Use a graph database to store the extracted entities and relationships as graph nodes and edges to obtain a knowledge graph.

[0048] S13: Use distributed storage and memory cache mechanism to store the knowledge graph.

[0049] Specifically, building and storing the knowledge graph is as follows:

[0050] S11 Data acquisition: The system acquires initial knowledge through a variety of data sources (such as public knowledge bases, user interaction records, etc.), uses large models to extract structured information from natural languages, and generates entity, relationship, and semantic aggregation triples.

[0051] S12 knowledge graph construction: Use a graph database (such as Neo4j) to store the extracted entities and relationships as graph nodes and edges, supporting efficient query and reasoning.

[0052] S13 storage solution: Use distributed storage and memory cache mechanisms (such as Redis) to improve system query efficiency.

[0053] In an exemplary embodiment, in step S2, it specifically includes:

[0054] S21: Get new information;

[0055] S22: When the new information is a new concept, the domain requirements have changed, or there is a knowledge conflict, the new information is subjected to validity judgment, conflict detection, and user feedback verification;

[0056] S23: After verification, the knowledge graph is updated to obtain an updated knowledge graph.

[0057] Specifically, the dynamic update of the knowledge graph is as follows:

[0058] S21 obtains new information and determines whether the update condition is triggered:

[0059] 1. New concept identification: Automatically determine whether the user input contains new concepts not recorded in the system through keyword matching, semantic similarity calculation or predefined domain term comparison. If a new concept is identified, the system marks it and prompts the administrator to confirm.

[0060] 2. Changes in domain requirements: New requirements proactively defined by business logic or domain experts, usually triggered by APIs or custom rules.

[0061] 3. Knowledge conflict detection: When the information entered by the user is inconsistent with the relationship or entity in the existing knowledge graph, the system detects the conflict through rule detection or anomaly recognition algorithm and triggers the update process.

[0062] S22 update process: conduct validity judgment, conflict detection and user feedback verification on new information to ensure the accuracy and consistency of map updates.

[0063] 1. Validity judgment: Ensure the validity of input information through credibility assessment of information sources (such as citing authoritative data sources) and data quality analysis (such as checking missing values ​​or format errors).

[0064] 2. Conflict detection: Use existing graph data to verify logical consistency. If a conflict is detected, the administrator is prompted to resolve it or select trusted data through automated policies.

[0065] 3. User feedback verification: Show update suggestions to users or administrators in related fields and collect confirmation information to ensure the reliability of new data.

[0066] As an optional implementation, the user intention understanding method based on the combination of a large model and a knowledge graph further includes:

[0067] Based on the updated knowledge graph, the semantic analysis model is trained with the user's natural language query text as input and the structured query statement or graph entity set as output to obtain a trained semantic analysis model; the semantic analysis model is BERT, GPT or a knowledge embedding network combined with the Transformer architecture.

[0068] Specifically, the updated feedback mechanism is as follows:

[0069] 1. Based on the updated knowledge graph, use existing semantic analysis models (such as BERT, GPT, etc.), or build custom models for specific fields (such as knowledge embedding networks combined with Transformer architecture).

[0070] 2. Input and output of the semantic analysis model: The input is the user's natural language query text, and the output is a structured query statement or a graph entity set.

[0071] 3. Model training process:

[0072] Data preparation: Extract training data from the knowledge graph, including entity, relationship labels and semantic descriptions.

[0073] Pre-trained model fine-tuning: Use transfer learning to adapt the general semantic model to the domain.

[0074] Verification and tuning: Use precise and standardized data sets to verify model performance and optimize parameters through cross-validation.

[0075] Update effect verification: Evaluate the accuracy and comprehensiveness of query results before and after the update through comparative analysis.

[0076] In an exemplary embodiment, in step S3, it specifically includes:

[0077] S31: Use the big model to perform entity recognition and relationship extraction on the updated knowledge graph to generate semantic knowledge.

[0078] S32: Based on the semantic knowledge, semantic similarity calculation is used to merge synonyms and remove redundant information to obtain a modified knowledge graph.

[0079] S33: According to the semantic consistency of the newly added nodes and relationships, the modified knowledge graph structure is adjusted to obtain the final knowledge graph.

[0080] Specifically, multi-source data fusion and deduplication are as follows:

[0081] S31 Semantic Extraction: Use a large model (LionButterfly trained by Callisto Technology) to perform entity recognition and relationship extraction on multi-source data to generate semantic knowledge.

[0082] S32 data deduplication and knowledge alignment:

[0083] 1. Semantic similarity calculation: Using the vector representation generated by the large model (LionButterfly trained by Callisto Technology), the semantic similarity is calculated by the following method:

[0084]

[0085] in, is the vector representation of data A, is the vector representation of data B, is the vector modulus length of data A, is the vector modulus of data B, and Sim(A, B) is the semantic similarity between data A and data B. This formula calculates the cosine value of the angle between two vectors. The closer the value is to 1, the higher the semantic similarity. The semantic similarity threshold is set according to the characteristics of the domain, usually between 0.8 and 0.9, to ensure the accuracy of high similarity matching.

[0086] 2. Synonym merging:

[0087] According to the semantic similarity results, if the similarity between two entities exceeds the threshold, they are judged as synonyms.

[0088] The merge operation includes: retaining the entity described by the core semantics as the main entity, merging the attributes and relationships of other synonym entities into the main entity, and updating the reference relationship in the graph.

[0089] 3. Redundant information removal:

[0090] For entities or relations with the same semantics but repeated, redundant items are located through semantic similarity matching.

[0091] Remove redundant entities or relationships to ensure that each node and edge in the graph is unique and has independent meaning.

[0092] S33 graph structure dynamic update:

[0093] Adjust the graph structure based on the semantic consistency of the newly added nodes and relationships:

[0094] 1. If the semantic similarity between the newly added entity and the existing node is high, the new node will not be added, but the relationship will be directed to the existing node.

[0095] 2. If the newly added relationship duplicates the existing relationship (the semantic similarity is higher than the threshold), the relationship weights are merged or the relationship attributes are updated.

[0096] 3. If the new information constitutes a new semantic path, the system dynamically adds nodes and edges to expand the coverage of the graph.

[0097] In an exemplary embodiment, in step S4, user intention analysis and deep reasoning are as follows:

[0098] S41 explicit demand analysis:

[0099] Criteria:

[0100] Explicit needs are intentions that users convey directly through explicit expression. The system uses the following criteria to determine:

[0101] Keyword matching: The system analyzes the core keywords, entities, and attributes in the user input. For example, in the sentence “recommend a computer suitable for students”, the keywords are “recommend”, “suitable for students”, and “computer”.

[0102] Syntactic structure analysis: Determine the semantic core word and its modifiers through the dependency syntax tree. For example, in the above sentence, "computer" is the core requirement and "suitable for students" is the modifier.

[0103] Knowledge graph query: Use the identified keywords and relationships to match relevant entities and relationships in the knowledge graph to verify the clarity of explicit requirements.

[0104] 2. Hidden demand mining:

[0105] Criteria:

[0106] Implicit needs are intentions that are not directly expressed by the user but can be derived through reasoning or context analysis. The system uses the following criteria to determine:

[0107] Contextual relevance: Analyze the relationship between the user's current input and historical interactions in context. For example, "recommended weekend places" may imply the user's preference for "relaxing activities" or "natural attractions."

[0108] Semantic completion: When user input is incomplete or ambiguous, the system uses entities and relationships in the knowledge graph to complete it. For example, when a user enters "places to play on weekends", it is inferred that "places to play" may be related to topics such as "travel" and "entertainment".

[0109] Preference inference: Combined with the user's historical behavior records, a large model is used to generate the user's possible interests and determine whether there are any unexpressed implicit needs.

[0110] 3. Semantic completion and disambiguation:

[0111] 1) How to determine whether a question contains implicit needs:

[0112] The system determines whether the user's question contains implicit requirements through the following methods:

[0113] Input ambiguity: When the user does not clearly express his / her intention (e.g. “recommend a good movie”), the system may judge his / her intention to include multiple possibilities.

[0114] Contextual conflict: When there is a logical difference between the current question and the context (e.g., the previous discussion was about "festival activities", but the question turned to "arrangements", which may imply specific needs that are not expressed).

[0115] Demand reasoning rules: The system has built-in reasoning rules to determine whether it is necessary to infer user needs based on different semantic scenarios (such as verb patterns such as "recommend" and "want").

[0116] 2) How are users’ past behavior preferences classified and saved?

[0117] Classification standards: User behaviors are classified and saved based on fields (such as "entertainment", "education", and "shopping"); within each field, they are further classified by attributes (such as time, frequency, and rating).

[0118] Storage method: Preference data is stored in a structured form in the user profile, and each preference category is associated with a specific label (such as "Movie preference: science fiction, rating > 4 stars").

[0119] Dynamic update: The system updates preference data based on user interactions in real time and interacts with the knowledge graph to add new tags and relationships.

[0120] 3) How to match and filter the user’s past preference information when obtaining implicit preferences:

[0121] Preference extraction: Extract historical preferences related to current needs from user portraits. For example, if the question involves "movie recommendation", the system extracts the user's "movie viewing history" and "rating preference". Multi-level matching:

[0122] Direct Match: The system directly matches preference data based on keywords or attributes.

[0123] Semantic reasoning matching: If the user does not directly express the attribute (such as "science fiction film"), the system determines the potential needs through semantic analysis and graph reasoning, and matches the historical records related to the current intention.

[0124] Sorting and filtering: Sort the matching results according to the timeliness and weight of user behavior. For example, recent preferences are given higher weights, while outdated preferences are gradually reduced in weight.

[0125] Through the above mechanism, the system can deeply analyze user intentions in scenarios with fuzzy input and implicit demand, and provide accurate and personalized responses.

[0126] In summary, this application achieves a deep understanding of user intent through the innovative integration of large models and knowledge graph technology, breaking through the shallow semantic limitations and static knowledge bottlenecks of existing technologies. This method is suitable for scenarios such as intelligent customer service, conversational AI, and recommendation systems, and can significantly improve user experience, with broad market value and application prospects.

[0127] The present application also provides an application scenario, which applies the above-mentioned user intent understanding method based on the combination of a large model and a knowledge graph. Specifically: The user intent understanding method based on the combination of a large model and a knowledge graph provided in this embodiment can be applied in artificial intelligence scenarios. The artificial intelligence scenario includes: knowledge graph construction and storage links, knowledge graph update links, multi-source data fusion and deduplication links, and user intent analysis and deep reasoning links; first, build and store the knowledge graph; when the update condition is triggered, dynamically update the knowledge graph to obtain an updated knowledge graph; perform multi-source data fusion and deduplication on the updated knowledge graph to obtain the final knowledge graph; based on the final knowledge graph, perform user intent analysis and deep reasoning on the data input by the user to obtain user needs.

[0128] The specific embodiments are as follows:

[0129] The user sends a query request to the system through natural language, "I want to find a laptop suitable for students." The system first extracts the two key entities of "students" and "laptops" through large model analysis, and then queries the knowledge graph to obtain relevant attributes, brand information, etc.

[0130] Combined with the user's historical behavior, the system infers that the user prefers cost-effective products and returns recommendations for laptops that meet the requirements.

[0131] Based on the same inventive concept, the embodiment of the present application also provides a user intention understanding system for implementing the user intention understanding method based on the combination of a large model and a knowledge graph. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the user intention understanding system based on the combination of a large model and a knowledge graph provided below can be found in the above limitations on the user intention understanding method based on the combination of a large model and a knowledge graph, and will not be repeated here.

[0132] In an exemplary embodiment, Figure 4 As shown, a user intention understanding system based on a combination of a large model and a knowledge graph is provided, and the user intention understanding system based on a combination of a large model and a knowledge graph includes:

[0133] The knowledge graph construction and storage module M1 is used to construct and store the knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: a public knowledge base and user interaction records.

[0134] The knowledge graph update module M2 is used to dynamically update the knowledge graph to obtain an updated knowledge graph when the update conditions are triggered; the update conditions include: new concept identification, domain demand changes and knowledge conflict detection.

[0135] The multi-source data fusion and deduplication module M3 is used to perform multi-source data fusion and deduplication on the updated knowledge graph to obtain the final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure.

[0136] The user intention analysis and deep reasoning module M4 is used to perform user intention analysis and deep reasoning on the data input by the user based on the final knowledge graph to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

[0137] This application breaks through the limitations of traditional keyword matching by combining knowledge graphs and large models, and realizes semantic mining of users' implicit needs. At the same time, the system supports automatic extraction and updating of knowledge, avoiding the static and inefficient maintenance of traditional knowledge bases.

[0138] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store knowledge graph data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a user intention understanding method based on a combination of a large model and a knowledge graph is implemented.

[0139] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0140] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the above-mentioned method embodiments are implemented when the processor executes the computer program.

[0141] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0142] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0143] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0144] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0145] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0146] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0147] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A user intention understanding method based on a combination of a large model and a knowledge graph, characterized in that: The user intention understanding method based on the combination of a large model and a knowledge graph includes: Construct and store a knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: a public knowledge base and user interaction records; When an update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph; the update condition includes: new concept identification, domain demand changes, and knowledge conflict detection; Perform multi-source data fusion and deduplication on the updated knowledge graph to obtain a final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure; Based on the final knowledge graph, user intention analysis and deep reasoning are performed on the data input by the user to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

2. The user intention understanding method based on the combination of a large model and a knowledge graph according to claim 1 is characterized in that: Build and store knowledge graphs, including: Based on multiple data sources, we acquire initial knowledge and use large models to extract structured information from natural language to generate entities, relationships, and semantic aggregation triples. Use a graph database to store the extracted entities and relationships as graph nodes and edges to obtain a knowledge graph; The knowledge graph is stored using distributed storage and memory cache mechanisms.

3. The user intention understanding method based on the combination of a large model and a knowledge graph according to claim 1 is characterized in that: When the update condition is triggered, the knowledge graph is dynamically updated to obtain an updated knowledge graph, which specifically includes: Get new information; When the new information is a new concept, the domain requirements have changed, or there is a knowledge conflict, the new information is subject to validity judgment, conflict detection, and user feedback verification; After the verification is completed, the knowledge graph is updated to obtain an updated knowledge graph.

4. The user intention understanding method based on the combination of a large model and a knowledge graph according to claim 3 is characterized in that: The user intention understanding method based on the combination of the large model and the knowledge graph also includes: Based on the updated knowledge graph, the semantic analysis model is trained with the user's natural language query text as input and the structured query statement or graph entity set as output to obtain a trained semantic analysis model; the semantic analysis model is BERT, GPT or a knowledge embedding network combined with the Transformer architecture.

5. The user intention understanding method based on the combination of a large model and a knowledge graph according to claim 1 is characterized in that: The updated knowledge graph is subjected to multi-source data fusion and deduplication to obtain the final knowledge graph, which specifically includes: Use the big model to perform entity recognition and relationship extraction on the updated knowledge graph to generate semantic knowledge; Based on the semantic knowledge, semantic similarity calculation is used to merge synonyms and remove redundant information to obtain a modified knowledge graph; According to the semantic consistency of the newly added nodes and relationships, the modified knowledge graph structure is adjusted to obtain the final knowledge graph.

6. The method for understanding user intention based on the combination of a large model and a knowledge graph according to claim 5 is characterized in that: The calculation formula of the semantic similarity is: in, is the vector representation of data A, is the vector representation of data B, is the vector modulus length of data A, is the vector modulus of data B, and Sim(A, B) is the semantic similarity between data A and data B.

7. A user intention understanding system based on a combination of a large model and a knowledge graph, characterized in that: The user intention understanding system based on the combination of a large model and a knowledge graph includes: A knowledge graph construction and storage module is used to construct and store a knowledge graph; the knowledge graph is a graph obtained by extracting structured information from multiple data sources based on a large model; the data sources include: a public knowledge base and user interaction records; A knowledge graph update module is used to dynamically update the knowledge graph to obtain an updated knowledge graph when an update condition is triggered; the update condition includes: new concept identification, domain demand changes, and knowledge conflict detection; A multi-source data fusion and deduplication module is used to perform multi-source data fusion and deduplication on the updated knowledge graph to obtain a final knowledge graph; the multi-source data fusion and deduplication include: semantic extraction, data deduplication and knowledge alignment, and dynamic update of graph structure; The user intention analysis and deep reasoning module is used to perform user intention analysis and deep reasoning on the data input by the user based on the final knowledge graph to obtain user needs; the user intention analysis and deep reasoning include: explicit demand analysis, implicit demand mining and semantic completion and disambiguation; the explicit demand is the intention directly conveyed by the user through clear expression; the implicit demand is the intention not directly expressed by the user but obtained through reasoning or context analysis; the semantic completion and disambiguation is to accurately understand the user's vague or implicit needs through semantic completion, context association and disambiguation processing.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the user intent understanding method based on the combination of a large model and a knowledge graph as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the user intention understanding method based on the combination of a large model and a knowledge graph as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the user intention understanding method based on the combination of a large model and a knowledge graph as described in any one of claims 1 to 6.

Images