Question clarification methods, apparatuses, and devices based on entity and intent ambiguity analysis

Abstract

The application discloses a question clarification method, device, equipment, medium and product based on entity and intention ambiguity analysis, relates to the field of intelligent question answering, and comprises the following steps: acquiring a question sentence of a user input; the following steps are cyclically executed until the entity ambiguity and intention ambiguity of the question sentence are less than or equal to a preset threshold: analyzing entities and intentions in the question sentence to obtain the entity ambiguity and intention ambiguity of the question sentence; in the case where the entity ambiguity and / or intention ambiguity is greater than the preset threshold, generating target clarification request content by using a large language model based on entity ambiguity corresponding to entity ambiguity and / or intention ambiguity corresponding to intention ambiguity; sending the target clarification request content to the user and acquiring target clarification response content of the user; and updating the semantic representation of the question sentence based on the target clarification response content. The application can dynamically quantify the semantic ambiguity of a user question by calculating the entity ambiguity and intention ambiguity.

Description

Technical Field

[0001] This application relates to the field of intelligent question answering technology, and in particular to a question clarification method, apparatus, device, medium and product based on entity and intent ambiguity analysis. Background Technology

[0002] Interactive applications such as intelligent question-answering systems and chatbots are widely used in customer service, information retrieval, and intelligent assistant scenarios. Users input questions in natural language, and the system needs to accurately understand the semantics and generate a response. However, natural language has inherent ambiguity and vagueness. For example, entity references are unclear (e.g., "apple" could refer to a fruit or a company), and intents are ambiguous (e.g., "turn on the device" does not specify the device type or operational details), making it difficult for the system to accurately interpret user needs, thus affecting the accuracy of the answer or triggering erroneous operations. Summary of the Invention

[0003] The purpose of this application is to provide a problem clarification method, apparatus, device, medium, and product based on entity and intent ambiguity analysis.

[0004] To achieve the above objectives, this application provides the following solution:

[0005] In a first aspect, this application provides a question clarification method based on entity and intent ambiguity analysis, comprising: acquiring a question statement input by a user; repeatedly executing the following steps until the entity ambiguity and intent ambiguity of the question statement are both less than or equal to a preset threshold: parsing the entities and intents in the question statement to obtain the entity ambiguity and intent ambiguity of the question statement; when the entity ambiguity and / or the intent ambiguity are greater than the preset threshold, using a large language model, generating target clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity; sending the target clarification request content to the user and acquiring the user's target clarification response content; and updating the semantic representation of the question statement based on the target clarification response content.

[0006] Secondly, this application provides a question clarification device based on entity and intent ambiguity analysis, comprising: an acquisition module for acquiring a question statement input by a user; a parsing module for repeatedly executing the following steps until the entity ambiguity and intent ambiguity of the question statement are both less than or equal to a preset threshold: parsing the entities and intents in the question statement to obtain the entity ambiguity and intent ambiguity of the question statement; a generation module for generating target clarification request content based on entity ambiguity corresponding to entity ambiguity and / or intent ambiguity corresponding to intent ambiguity when the entity ambiguity and / or intent ambiguity are greater than the preset threshold, using a large language model; a sending module for sending the target clarification request content to the user and acquiring the user's target clarification response content; and an updating module for updating the semantic representation of the question statement based on the target clarification response content.

[0007] Thirdly, this application provides a computer device, including: 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 steps of the problem clarification method based on entity and intent ambiguity analysis as described above.

[0008] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the problem clarification method based on entity and intent ambiguity analysis as described above.

[0009] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the problem clarification method based on entity and intent ambiguity analysis as described above.

[0010] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0011] This application provides a question clarification method, apparatus, device, medium, and product based on entity and intent ambiguity analysis. By calculating entity ambiguity and intent ambiguity, the semantic ambiguity of user questions is transformed into quantifiable indicators, avoiding blind clarification. By generating targeted clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity, targeted clarification is achieved, reducing redundant follow-up questions. Natural language clarification improves user cooperation and clarification efficiency. Through a closed-loop process of "parsing → clarification → updating," the semantic representation of the question is continuously iterated, ensuring the cumulative optimization of semantic understanding in multi-turn dialogues, thereby more accurately parsing user needs and improving the accuracy of answers. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis provided in an embodiment of this application;

[0014] Figure 2 A schematic diagram of an interactive clarification tool provided in an embodiment of this application;

[0015] Figure 3 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis, provided for another embodiment of this application;

[0016] Figure 4 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis, provided as another embodiment of this application;

[0017] Figure 5 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis, provided in yet another embodiment of this application;

[0018] Figure 6 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis is provided in another embodiment of this application;

[0019] Figure 7 A flowchart illustrating a problem clarification method based on entity and intent ambiguity analysis is provided in another embodiment of this application;

[0020] Figure 8 A schematic diagram of the functional modules of a problem clarification device based on entity and intent ambiguity analysis provided in an embodiment of this application;

[0021] Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] In one exemplary embodiment, such as Figure 1 As shown, a question clarification method based on entity and intent ambiguity analysis is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, it includes steps 102 to 110, whereby the question-answering agent can execute steps 102 to 110. This question-answering agent can run as a core component of an intelligent question-answering system mounted on a computer device. Wherein:

[0025] Step 102: Obtain the question statement input by the user;

[0026] The question-answering agent includes a Large Language Model (LLM), which integrates interactive clarification tools. Figure 2 This is a schematic diagram of the interactive clarification tool proposed in the embodiments of this application, used to facilitate direct user interaction. Users can input question statements through the interactive clarification tool. The question statements can be inquiries actively raised by the user through text, voice, or other means, such as... Figure 2 As shown, the question input by the user can be "What made Zhang San famous?".

[0027] Repeat steps 104 to 110 until both the entity ambiguity and intent ambiguity of the question statement are less than or equal to a preset threshold:

[0028] Step 104: Parse the entities and intents in the question statement to obtain the entity ambiguity and intent ambiguity of the question statement;

[0029] Wherein, entity ambiguity refers to the degree of ambiguity of entities in the question statement, and intent ambiguity refers to the degree of ambiguity of intent in the question statement.

[0030] Step 106: When the entity ambiguity and / or the intent ambiguity are greater than the preset threshold, use a large language model to generate target clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity;

[0031] The target clarification request content can be a clarification request generated by the interactive clarification tool when there is doubt, misunderstanding or ambiguity in the content expressed by the user, based on entity ambiguity and / or question ambiguity in the question statement, requiring the user to further explain, supplement details or clarify the meaning.

[0032] Step 108: Send the target clarification request content to the user and obtain the target clarification response content from the user;

[0033] The target clarification response content is the user's answer or feedback, which is supplementary information or explicit explanation provided in response to the target clarification request content.

[0034] Step 110: Update the semantic representation of the question statement based on the target clarification response content.

[0035] By updating the semantic representation of the question statement, the semantic representation of the question statement can be made clearer, allowing the question-answering agent to understand user needs more accurately.

[0036] In this embodiment, by calculating entity ambiguity and intent ambiguity, the semantic ambiguity of user questions is transformed into quantifiable indicators, avoiding blind clarification. Targeted clarification request content is generated based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity, achieving targeted clarification, reducing redundant follow-up questions, and improving user cooperation and clarification efficiency through natural language clarification. Through a closed-loop process of "parsing → clarification → updating," the semantic representation of the question is continuously iterated, ensuring cumulative optimization of semantic understanding in multi-turn dialogues, thereby enabling more accurate parsing of user needs and improving the accuracy of responses.

[0037] In another exemplary embodiment of this application, such as Figure 3 As shown, step 104 above can be replaced by steps 1041 to 1043:

[0038] Step 1041: Decompose the question statement into at least one question triple, each question triple including a head entity, a predicate, and a tail entity, the head entity and the tail entity are collectively referred to as entities;

[0039] Each question statement can be decomposed, decoupling complex question statements into multiple simple triples; in each round T (round index), the intelligent question answering system will follow the guidelines. and interaction history Generate an action, interaction history The action can be represented by the following formulas (1) and (2):

[0040] (1);

[0041] (2);

[0042] in, Inst consists of pre-written tool descriptions, tool usage methods, and interaction formats. E represents a set of examples, such as question-and-answer examples and tool call examples. Q represents question-and-answer statements. This indicates the action generated in the first round. This represents the results of the first round of observations, i.e., the actions performed. The feedback result obtained is defined as follows: , This indicates the target of the first round of user clarification in the response content; similarly, This represents the action generated in round T. This represents the observation results from round T, i.e., the actions performed. The feedback result obtained is defined as follows: , This indicates the target clarification response content for users in round T; This represents the action generated in round T+1.

[0043] This can be broken down into querying information related to the entity and intent for each triple.

[0044] Step 1042: For each head entity or tail entity in the aforementioned problem triple, invoke the entity query tool to query the first information related to the head entity or tail entity in the knowledge base, and generate entity fuzziness based on the first information;

[0045] The question-answering agent also integrates an entity query tool. This tool can search for entity nodes in the knowledge base using a given surface name and return the formal entity name and its distinguishing features (such as description and type). The relationship query tool can include a SearchNodes tool for retrieving entities. For example, the SearchNodes tool can be used to retrieve node information in the knowledge base that is similar to or related to the input entity surface name, analyze entity ambiguity, and derive entity fuzziness.

[0046] like Figure 2 As shown, after obtaining the question "What is Zhang San famous for?", the SearchNodes tool can be used to retrieve the first information related to the entity in the knowledge base, and the entity fuzziness can be calculated based on the first information.

[0047] Step 1043: Based on the head entity and predicate in the at least one question triple, invoke the predicate analysis tool to query the knowledge base for second information related to the head entity and predicate, and generate intent ambiguity based on the second information;

[0048] The question-answering agent also integrates a predicate analysis tool, also known as a relation query tool. This tool can identify and rank semantically relevant relation predicates by querying a one-hop subgraph around a specified entity. It supports the parsing of complex structures (such as CVT nodes). The predicate analysis tool may include a SearchGraphPattern tool for retrieving predicates. For example, the SearchGraphPattern tool can be used to retrieve knowledge graphs (such as one-hop subgraphs) in the knowledge base to analyze the semantic ambiguity of entity relationships and derive intent ambiguity.

[0049] like Figure 2 As shown, the SearchGraphPattern tool can be used to retrieve the first information related to the predicate in the knowledge base, and the intent ambiguity can be calculated based on the second information.

[0050] Step 106, "using a large language model, to generate target clarification request content based on entity ambiguity corresponding to the entity ambiguity and / or intent ambiguity corresponding to the intent ambiguity," includes:

[0051] The interactive clarification tool in the large language model is invoked to generate the target clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity.

[0052] Among them, the interactive clarification tool, also known as the fuzzy clarification plugin, is a disambiguation management tool. The interactive clarification tool can be the AskForClarification tool, which can request clarification from the user when needed. When the entity ambiguity and / or the intent ambiguity are greater than the preset threshold, it indicates that entity ambiguity and / or intent ambiguity has been detected. At this time, the AskForClarification tool can be triggered to automatically generate a natural language clarification request (i.e., the target clarification request content), interrupt the interactive process, request clarification from the user, and continue processing after the user inputs feedback.

[0053] For each interaction, it's necessary to determine whether clarification of the entity or intent is needed. If so, the interaction clarification tool should be invoked to allow the user to clarify the issue. For example... Figure 2 As shown, when the entity ambiguity is greater than a preset threshold, the target clarification request content "Which Zhang San do you want to ask about? Is he an American athlete or a Chinese writer?" can be generated and displayed, and the user's target clarification response content "I want to ask about a Chinese writer" can be obtained; similarly, when the intent ambiguity is greater than a preset threshold, the target clarification request content "Do you want to know about her works, awards, or profession?" can be generated and displayed, and the user's target clarification response content "I want to know about his profession" can be obtained.

[0054] In this embodiment, by decomposing the question statement into question triples, entity query tools and predicate analysis tools are used to quantify the ambiguity of entities and intents respectively, providing accurate basis for question clarification; by generating target clarification requests with the help of interactive clarification tools, a closed loop from fuzzy identification to precise questioning is realized, which can effectively improve the accuracy of semantic understanding, reduce the cost of repeated communication for users, and significantly improve interaction efficiency and service quality in scenarios such as intelligent question answering and customer service.

[0055] In another exemplary embodiment of this application, such as Figure 4 As shown, the step 1042 above, "generating entity ambiguity based on the first information", can be replaced by the following steps 10421 to 10425:

[0056] Step 10421: Based on the popularity of each entity in the first information, determine the prior probability of the corresponding entity;

[0057] Among these, Bayes' theorem can be used to calculate the posterior probability of an entity. To evaluate the ambiguity of entities retrieved by the SearchNodes tool relative to a given question statement Q, a Bayesian-based metric was designed to determine... Whether the entities in the text are ambiguous, the goal is to compute First, each entity can be used. Popularity in knowledge bases (such as Freebase) To calculate its prior probability, entity The prior probability can be expressed by the following formula (3):

[0058] (3);

[0059] in, It represents the sum of the popularity of N entities. The popularity of an entity in the knowledge base can refer to the frequency and extent to which the entity is mentioned, used, or associated in the knowledge base. It is a key indicator for measuring the importance and attention of an entity in the knowledge system.

[0060] Step 10422: Based on the description information of each entity in the first information and the question statement, determine the conditional probability of the question statement under the condition that the corresponding entity exists;

[0061] Among them, conditional probability Entities can be measured With question statement The correlation between them can be determined by... Description Conditional probability derived using LLM. It can be expressed by the following formula (4):

[0062] (4);

[0063] Where PPL(·) represents the ambiguity function, which can be normalized using the softmax function to ensure the validity of the probability distribution. It can be an entity The semantic information is structured or textualized to accurately express the entity's connotation, attributes, relationships and other key features.

[0064] Step 10423: Based on the prior probability and conditional probability of each entity, determine the posterior probability of the corresponding entity using Bayes' theorem;

[0065] Among these, Bayes' theorem can be used to calculate the posterior probability of an entity. As shown in the following formula (5):

[0066] (5);

[0067] Step 10424: Determine the entropy of the distribution of multiple entities based on the posterior probability of each entity;

[0068] The entropy of multiple entity distributions can be expressed by the following formula (6):

[0069] (6);

[0070] in yes Softmax regularization over all entities is used to quantify the uncertainty or ambiguity of the retrieved entities.

[0071] Step 10425: Generate entity ambiguity based on the entropy of the distribution of the multiple entities.

[0072] The fuzziness score of an entity (i.e., entity fuzziness) can be calculated by dividing H by the maximum possible entropy. We obtain N, which is the total number of entities retrieved. logN represents the maximum level of confusion that may occur when the number of entities is N. The entity ambiguity can be calculated using the following formula (7). .

[0073] (7);

[0074] The entity ambiguity ranges from 0 to 1. A score close to 0 indicates low ambiguity (a single entity clearly matches question Q), while a score close to 1 indicates high ambiguity (multiple entities have similar relevance) and requires clarification.

[0075] In this embodiment, by introducing Bayes' theorem, the popularity of entities, descriptive information, and user question statements are combined to scientifically calculate the prior probability, conditional probability, and posterior probability of entities, and then the entity ambiguity is quantified based on the entropy of the probability distribution. This approach provides a data-driven and accurate measurement method for identifying entity ambiguity, which is more objective and interpretable, and can effectively improve the accuracy of intelligent question answering systems in understanding the semantics of entities in user questions, thereby optimizing question clarification and interactive experience.

[0076] In another exemplary embodiment of this application, such as Figure 5 As shown, the step 1043 above, "generating intent ambiguity based on the second information," can be replaced by the following steps 10431 to 10437:

[0077] Step 10431: Determine the candidate tail entities in the second information that match the head entity and predicate in each of the question triples;

[0078] The head entity can be represented as The predicate can be represented as The candidate tail entity can be represented as .

[0079] Step 10432: Generate candidate triples based on the head entity, predicate, and candidate tail entity in each of the aforementioned problem triples;

[0080] The candidate triplet can be represented as .

[0081] Step 10433: Determine the joint prior probability of the candidate triples;

[0082] The joint prior probability can be expressed as: Prior probability Capturing these components is helpful for explanation The importance of their joint efforts.

[0083] Step 10434: Given a candidate triple consisting of a head entity, a predicate, and a candidate tail entity, determine the semantic matching probability between the question statement and the corresponding candidate triple, wherein the probability is used to measure the explanatory power of the corresponding candidate triple for the question statement.

[0084] The semantic matching probability can be expressed as: , indicating that in a given predicate Its tail entity and predicting entities In the case of Q, the likelihood value of Q.

[0085] Step 10435: Based on the joint prior probability and semantic matching probability, determine the posterior probability of each predicate using Bayes' theorem;

[0086] Among these, Bayes' theorem can be used to calculate the posterior probability of the predicate. To assess the ambiguity of predicates retrieved using the SearchGraphPattern tool, a metric can be used to evaluate the uncertainty of the relationship between the predicted entity e′ and the question statement Q. Let... Represents a list of candidate predicate-tail entity pairs, where It is a predicate. It is the corresponding tail entity sampled from the knowledge graph, and the goal is to compute... It can be simplified by Bayes' theorem to the following formula (8):

[0087] (8);

[0088] Step 10436: Determine the entropy of the distribution of multiple predicates based on the posterior probability of each predicate;

[0089] The entropy of multiple predicate distributions can be expressed as: ,in yes Softmax regularization for all predicates.

[0090] Step 10437: Generate intent ambiguity based on the entropy of the distribution of the multiple predicates.

[0091] Wherein, following the same ambiguity-based method as for entities, the ambiguity score of the predicate (i.e., intent, ambiguity) can be expressed by the following formula (9):

[0092] (9);

[0093] Where M is the total number of entities retrieved, and logM represents the maximum level of confusion that may occur when the number of entities is M. Similar to the fuzziness score of entities, a high fuzziness score for predicates indicates that multiple predicates exhibit similar relevance and require further clarification.

[0094] This application embodiment designs a plug-and-play plugin that uses Bayesian inference to dynamically calculate entity ambiguity scores and predicate ambiguity scores during multi-round interactions based on the tool execution results, thereby achieving a quantitative assessment of the degree of ambiguity.

[0095] In this embodiment, candidate triples are constructed, and the joint prior probability and semantic matching probability are fused using Bayes' theorem to calculate the posterior probability of the predicate. Then, the intent ambiguity is accurately quantified using the probability distribution entropy. This method provides a data-driven and measurable analytical path for intent ambiguity identification, offering greater scientific rigor and accuracy. It effectively helps intelligent question-answering systems understand the user's true intent, optimize question clarification strategies, and improve the accuracy and efficiency of human-computer interaction.

[0096] In another exemplary embodiment of this application, such as Figure 6 As shown, step 106, "using a large language model, based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity, to generate target clarification request content," includes:

[0097] Step 1061: Invoke the interactive clarification tool of the large language model, and generate the first clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity of the intent ambiguity;

[0098] When the question pertains to a legally related insurance claim, it could be, "Does the 'force majeure' clause in this property insurance contract apply to losses caused by extreme weather?" The substantive ambiguity in this question includes the lack of a specific type of "contract" (e.g., home insurance, vehicle insurance, business property insurance), and the ambiguity of intent, including the potential for different definitions of "extreme weather" in the insurance terms: strictly conforming to the disaster types listed in the contract (e.g., typhoons, floods, earthquakes); or falling under a catch-all clause of "other force majeure events," requiring judgment based on insurance industry practices. Therefore, the initial clarification request could be: "What type of insurance contract is this? (e.g., home insurance, vehicle insurance), and does the 'force majeure' clause in the contract explicitly list 'extreme weather' situations (e.g., typhoons, torrential rains, floods)?"

[0099] When the question statement pertains to smart home appliance control commands related to smart homes, the question statement could be something like, "Help me turn on the living room devices." The entity ambiguity in this question statement includes the unclear referencing of "living room devices," as there may be multiple devices in the living room such as a TV, air conditioner, lights, and air purifier, and it's unclear which one the user specifically wants to turn on. The intent ambiguity includes the fact that the action of "turning on" can have different meanings for different devices. For example, turning on the TV might mean turning on the power and switching to the last viewed channel; turning on the air conditioner might involve selecting cooling or heating modes and setting the temperature; turning on the lights might require specifying whether all lights are on or only lights in a specific area are on. Therefore, the generated first clarification request could be, "Hello, which device in the living room would you like to turn on? Is it the TV, air conditioner, lights, or another device? Also, are there any specific settings requirements for the device you want to turn on, such as changing TV channels, setting the air conditioner's mode and temperature, or adjusting the light brightness?"

[0100] Step 1062: Simulate a response to the first clarification request content using the user simulator of the large language model, and generate simulated clarification response content;

[0101] The large language model also integrates a user simulator. For insurance claims, the simulated clarification response generated by the user simulator can include multiple scenarios, such as: "It's corporate property insurance, and the terms list 'heavy rain' and 'flood'.", "Vehicle insurance, the terms only mention 'natural disasters,' without specifying the weather.", "What constitutes 'extreme weather'? We had a once-in-a-decade snowfall here, and the insurance agent said it doesn't count."

[0102] Regarding the issue of smart home appliance control commands, the simulated clarification response generated by the user simulator could be something like, "I want to turn on the living room TV and switch to the news channel," or "Turn on the living room air conditioner, cooling mode, 26 degrees," or "What do you mean by specific setting requirements? I just want to turn on the lights."

[0103] Step 1063: Based on the simulated clarification response content, generate the target clarification request content.

[0104] Regarding insurance claims, analysis of simulated clarification responses can reveal that "extreme weather" may lack quantifiable standards in the contract (such as rainfall, snow depth, etc.), requiring users to provide details of the terms. The optimized target clarification request could be something like, "What type of insurance contract is it? (e.g., house, vehicle, business property); Does the 'force majeure' clause in the contract clearly list specific types of 'extreme weather' (e.g., typhoon, torrential rain, flood)? If 'natural disaster' is mentioned, is there a definition of 'extreme' (e.g., 'rainfall exceeding 500mm / day', 'snow depth exceeding 30cm')?"

[0105] Regarding the issue of smart home appliance control commands, analysis of the simulated clarification response reveals that the phrase "specific setting requirements" is rather technical and may be difficult for some users to understand. A more easily understood alternative is needed. The optimized target clarification request could be something like, "Which device in your living room would you like to turn on? For example, the TV, air conditioner, lights, or other devices? Are there any specific operational requirements for the device you choose to turn on? For example, which channel should the TV be switched to, whether the air conditioner should be set to cooling, heating, or another mode, what temperature should be set, and whether all lights should be turned on, or adjusted in brightness or dimming, etc."

[0106] It should be noted that, in other embodiments, the target clarification request content can also be generated directly based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity of the intent ambiguity.

[0107] In some embodiments, the intelligent question-answering system may include a dual-agent interaction framework, through which a user simulator driven by a language model can interact with the question-answering agent, adaptively optimizing the generation of logical forms, thereby gradually eliminating the influence of ambiguity and simulating the user's response to the clarification request issued by the question-answering agent. Given the golden SPARQL query S′ and the content of the first clarification request, the agent generates the simulated clarification response content in the t-th round, as shown in the following formula (10):

[0108] (10);

[0109] By integrating information from multiple dimensions, the answer to the complex question is finally generated.

[0110] In this embodiment, a first clarification request content is generated by combining entity and intent ambiguity. A user simulator is used to rehearse the user response, and the target clarification request content is optimized based on the simulation results (simulated clarification response content). This approach can identify potential problems in the clarification request in advance (such as obscure wording or ambiguous options), improve the accuracy and user acceptance of the clarification content, reduce multiple rounds of invalid interactions, and effectively improve dialogue efficiency and user experience in scenarios such as intelligent question answering and customer service.

[0111] This application's embodiments relate to the field of knowledge graph question answering (KGQA) technology in the legal field. While legal questions are mostly expressed in a colloquial context, legal terminology is highly specialized, making direct text matching or vector matching ineffective. This application's embodiments, in the process of parsing natural language queries into executable logical expressions, analyze and clarify the ambiguity of the question's entity and intent to efficiently retrieve answers from a structured knowledge graph. Furthermore, by optimizing the generation of the target clarification request content based on simulation results (simulated clarification response content), the matching effect is excellent.

[0112] In another exemplary embodiment of this application, such as Figure 7 As shown, the problem clarification method based on entity and intent ambiguity analysis further includes the following steps:

[0113] Step 111: If both the entity ambiguity and the intent ambiguity are less than or equal to the preset threshold, invoke the SPARQL execution tool to execute a SPARQL query based on the logical structure generated by the semantic representation.

[0114] The large language model integrates a SPARQL execution tool. When both entity ambiguity and intent ambiguity are less than or equal to a preset threshold, it indicates that the semantic representation of the question statement is clear. The SPARQL execution tool can directly perform SPARQL queries on the knowledge base, providing basic query functionality. The Question Answering Agent (QA Agent) receives the user's target clarification response, updates the semantic representation of the question statement, executes the constrained query using the SPARQL tool, and updates the logical form based on Bayesian inference results. The user simulator (driven by LLMs) simulates possible feedback paths to further constrain the query conditions. The final generated SPARQL query is executed after verification to be unambiguous. The system returns the results and can output the reasoning process and confidence assessment. Experiments show that this method improves accuracy by 3.5% on complex query datasets (such as CWQ), especially in scenarios with limited samples.

[0115] In this embodiment, by setting a threshold for the ambiguity of entities and intents, the SPARQL execution tool is invoked only when the semantics are clear. The query is executed based on the logical structure of the semantic representation, which avoids query errors caused by ambiguity, ensures the accuracy of SPARQL queries, improves the efficiency of knowledge graph retrieval, and reduces the consumption of invalid query resources. This enables efficient and accurate semantic understanding and execution in scenarios such as intelligent question answering and data retrieval.

[0116] This application aims to address the ambiguity problem in Knowledge Graph Question Answering (KGQA). The method of this application dynamically handles entity and intent ambiguities by introducing a fuzzy clarification plugin into interactive knowledge graph question answering. Firstly, a novel fuzzy clarification plugin is proposed to dynamically handle ambiguity through interactive clarification. A Bayesian inference mechanism is employed to quantify query ambiguity and guide the LLM (Local Language Management Model) to determine when and how to request clarification from the user within a multi-turn dialogue framework. Secondly, a dual-agent interaction framework is developed, in which a user simulator based on a large language model can iteratively refine the logical form by simulating user feedback.

[0117] Based on the same inventive concept, this application also provides a problem clarification device based on entity and intent ambiguity analysis for implementing the problem clarification method based on entity and intent ambiguity analysis described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the problem clarification device based on entity and intent ambiguity analysis provided below can be found in the limitations of the problem clarification method based on entity and intent ambiguity analysis above, and will not be repeated here.

[0118] In one exemplary embodiment, such as Figure 8 As shown, a problem clarification device 200 based on entity and intent ambiguity analysis is provided, comprising:

[0119] Module 201 is used to obtain the question statement input by the user;

[0120] The parsing module 202 is used to repeatedly execute the following steps until the entity ambiguity and intent ambiguity of the question statement are both less than or equal to a preset threshold: parsing the entities and intents in the question statement to obtain the entity ambiguity and intent ambiguity of the question statement;

[0121] The generation module 203 is used to generate target clarification request content based on entity ambiguity corresponding to entity ambiguity and / or intent ambiguity corresponding to intent ambiguity when the entity ambiguity and / or intent ambiguity are greater than the preset threshold, using a large language model.

[0122] The sending module 204 is used to send the target clarification request content to the user and obtain the target clarification response content from the user;

[0123] The update module 205 is used to update the semantic representation of the question statement based on the target clarification response content.

[0124] As an optional implementation, the parsing module 202 includes: a decomposition submodule, used to decompose the question statement into at least one question triple, each of the question triples including a head entity, a predicate, and a tail entity, the head entity and the tail entity being collectively referred to as entities; a first generation submodule, used to, for each of the question triples, invoke an entity query tool to query first information related to the head entity or tail entity in a knowledge base, and generate entity ambiguity based on the first information; a second generation submodule, used to, based on the head entity and predicate in the at least one question triple, invoke a predicate analysis tool to query second information related to the head entity and predicate in the knowledge base, and generate intent ambiguity based on the second information; the generation module includes: a third generation submodule, used to invoke an interaction clarification tool in a large language model to generate target clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity.

[0125] As an optional implementation, the first generation submodule includes: a first determining unit, configured to determine the prior probability of a corresponding entity based on the popularity of each entity in the first information; a second determining unit, configured to determine the conditional probability of a question statement given the existence of the corresponding entity based on the description information of each entity in the first information and the question statement; a third determining unit, configured to determine the posterior probability of a corresponding entity using Bayes' theorem based on the prior and conditional probabilities of each entity; a fourth determining unit, configured to determine the entropy of multiple entity distributions based on the posterior probability of each entity; and a first generation unit, configured to generate entity ambiguity based on the entropy of the multiple entity distributions.

[0126] As an optional implementation, the second generation submodule includes: a fifth determining unit, configured to determine candidate tail entities in the second information that match the head entity and predicate in each of the question triples; a second generation unit, configured to generate candidate triples based on the head entity, predicate, and candidate tail entities in each of the question triples; a sixth determining unit, configured to determine the joint prior probability of the candidate triples; a seventh determining unit, configured to determine the semantic matching probability between the question statement and the corresponding candidate triples, given a candidate triple consisting of a head entity, a predicate, and a candidate tail entity, wherein the probability is used to measure the interpretability of the corresponding candidate triples for the question statement; an eighth determining unit, configured to determine the posterior probability of each predicate using a Bayesian formula based on the joint prior probability and the semantic matching probability; a ninth determining unit, configured to determine the entropy of multiple predicate distributions based on the posterior probability of each predicate; and a third generation unit, configured to generate intent ambiguity based on the entropy of the multiple predicate distributions.

[0127] As an optional implementation, the third generation module includes: a fourth generation submodule, used to call the interactive clarification tool of the large language model to generate a first clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity of the intent ambiguity; a fifth generation submodule, used to simulate a response to the first clarification request content through the user simulator of the large language model to generate simulated clarification response content; and a sixth generation submodule, used to generate target clarification request content based on the simulated clarification response content.

[0128] As an optional implementation, the device further includes: a calling module, configured to call a SPARQL execution tool to perform a SPARQL query based on the logical structure generated by the semantic representation when both the entity ambiguity and the intent ambiguity are less than or equal to the preset threshold.

[0129] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 9As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a problem clarification method based on entity and intent fuzzy analysis.

[0130] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0131] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0132] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0133] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0134] 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, data stored, data displayed, 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 the relevant data must comply with relevant regulations.

[0135] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this 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 can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0136] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0137] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0138] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A problem clarification method based on entity and intent ambiguity analysis, characterized in that, The problem clarification method based on entity and intent ambiguity analysis includes: Get the question statement input by the user; Repeat the following steps until both the entity ambiguity and intent ambiguity of the question statement are less than or equal to a preset threshold: The question statement is decomposed into at least one question triple, and each question triple includes a head entity, a predicate, and a tail entity. The head entity and the tail entity are collectively referred to as entities. For each of the head or tail entities in the aforementioned problem triples, an entity query tool is invoked to query the knowledge base for first information related to the head or tail entity, and based on the first information, entity fuzziness is generated. Based on the head entity and predicate in the at least one question triple, a predicate analysis tool is invoked to query second information related to the head entity and predicate in the knowledge base, and based on the second information, intent ambiguity is generated; When the entity ambiguity and / or the intent ambiguity are greater than the preset threshold, a target clarification request content is generated using a large language model based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity. Send the target clarification request content to the user and obtain the target clarification response content from the user; Based on the clarified response content, the semantic representation of the question statement is updated.

2. The problem clarification method based on entity and intent ambiguity analysis according to claim 1, characterized in that, The step of using a large language model to generate target clarification request content based on entity ambiguity corresponding to entity ambiguity and / or intent ambiguity corresponding to intent ambiguity includes: The interactive clarification tool in the large language model is invoked to generate the target clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity corresponding to the intent ambiguity.

3. The problem clarification method based on entity and intent ambiguity analysis according to claim 2, characterized in that, The step of generating entity ambiguity based on the first information includes: Based on the popularity of each entity in the first information, the prior probability of the corresponding entity is determined. Based on the description information of each entity in the first information and the question statement, determine the conditional probability of the question statement under the condition that the corresponding entity exists; Based on the prior probability and conditional probability of each entity, the posterior probability of the corresponding entity is determined by Bayes' theorem. Based on the posterior probability of each entity, the entropy of the distribution of multiple entities is determined; Based on the entropy of the distribution of the multiple entities, entity ambiguity is generated.

4. The problem clarification method based on entity and intent ambiguity analysis according to claim 2, characterized in that, The step of generating intent ambiguity based on the second information includes: Identify candidate tail entities in the second information that match the head entity and predicate in each of the problem triples; Based on the head entity, predicate, and candidate tail entity in each of the aforementioned problem triples, candidate triples are generated; Determine the joint prior probability of the candidate triples; Given a candidate triple consisting of a head entity, a predicate, and a candidate tail entity, determine the semantic matching probability between the question statement and the corresponding candidate triple, whereby the probability is used to measure the explanatory power of the corresponding candidate triple for the question statement. Based on the joint prior probability and semantic matching probability, the posterior probability of each predicate is determined by Bayes' theorem. Based on the posterior probability of each predicate, determine the entropy of the distribution of multiple predicates; Based on the entropy of the distribution of the multiple predicates, intent ambiguity is generated.

5. The problem clarification method based on entity and intent ambiguity analysis according to claim 1, characterized in that, The step of generating target clarification request content using a large language model, based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity of the intent ambiguity, includes: The interactive clarification tool of the large language model is invoked to generate the first clarification request content based on the entity ambiguity corresponding to the entity ambiguity and / or the intent ambiguity of the intent ambiguity. The user simulator of the large language model simulates the response to the first clarification request, generating simulated clarification response content. Based on the simulated clarification response content, the target clarification request content is generated.

6. The problem clarification method based on entity and intent ambiguity analysis according to any one of claims 1 to 5, characterized in that, The problem clarification method based on entity and intent ambiguity analysis also includes: If both the entity ambiguity and the intent ambiguity are less than or equal to the preset threshold, the SPARQL execution tool is invoked to perform a SPARQL query based on the logical structure generated by the semantic representation.

7. A problem clarification device based on entity and intent ambiguity analysis, characterized in that, The question clarification device based on entity and intent ambiguity analysis includes: The acquisition module is used to acquire the question statement input by the user; The parsing module is used to repeatedly execute the following steps until both the entity ambiguity and intent ambiguity of the question statement are less than or equal to a preset threshold: The question statement is decomposed into at least one question triple, and each question triple includes a head entity, a predicate, and a tail entity. The head entity and the tail entity are collectively referred to as entities. For each of the head or tail entities in the aforementioned problem triples, an entity query tool is invoked to query the knowledge base for first information related to the head or tail entity, and based on the first information, entity fuzziness is generated. Based on the head entity and predicate in the at least one question triple, a predicate analysis tool is invoked to query second information related to the head entity and predicate in the knowledge base, and based on the second information, intent ambiguity is generated; The generation module is used to generate target clarification request content based on entity ambiguity corresponding to entity ambiguity and / or intent ambiguity corresponding to intent ambiguity when the entity ambiguity and / or intent ambiguity are greater than the preset threshold, using a large language model. The sending module is used to send the target clarification request content to the user and obtain the target clarification response content from the user; The update module is used to update the semantic representation of the question statement based on the target clarification response content.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the problem clarification method based on entity and intent ambiguity analysis as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the problem clarification method based on entity and intent ambiguity analysis as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the problem clarification method based on entity and intent ambiguity analysis as described in any one of claims 1-6.

Images