Knowledge graph-based dynamic progressive consultation guiding method and system
By using a knowledge graph-based dynamic progressive consultation guidance method, the problems of lagging user consultation tendency capture and rigid guidance strategies are solved. This enables accurate positioning and flexible guidance of users' core needs, thereby improving consultation efficiency and user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-04-07
Smart Images
Figure CN121809670A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of consultation content analysis, specifically to a dynamic progressive consultation guidance method and system based on knowledge graphs. Background Technology
[0002] User needs in consulting service scenarios are becoming increasingly diverse and complex. Multi-round dialogues have become the mainstream form of consulting. Accurately analyzing the entity relationships in the consulting content, efficiently identifying the core needs of users, and providing targeted guidance are directly related to service response efficiency and user experience.
[0003] Current consultation content analysis mostly relies on keyword matching or fixed rule bases, making it difficult to deeply explore entity relationships and semantic logic in multi-turn dialogues. Especially when consultation scenarios involve multiple entities or hidden obstacles to needs, the potential relationships between entities are easily overlooked, the capture of user consultation tendencies is lagging, the guidance strategy is rigid, it is difficult to efficiently overcome obstacles in consultation, and it cannot meet the needs for accurate and flexible analysis in complex consultation scenarios.
[0004] In summary, existing technologies suffer from technical problems such as a lag in capturing user consultation tendencies, rigid guidance strategies, and an inability to accurately pinpoint users' core needs. Summary of the Invention
[0005] This application provides a dynamic, progressive consultation guidance method and system based on knowledge graphs, aiming to solve the technical problems in existing technologies such as lagging capture of user consultation tendencies, rigid guidance strategies, and inability to accurately locate users' core needs.
[0006] In view of the above problems, the technical solution to achieve the present application is as follows: In a first aspect, this application provides a dynamic progressive consultation guidance method based on a knowledge graph. The method includes: marking entities in a consultation scenario using multi-turn consultation dialogues; extracting entity relationships from consultation content information corresponding to the multi-turn dialogues; setting up the knowledge graph based on the entities and entity relationships in the consultation scenario; evaluating goal achievement and identifying obstacle nodes in the knowledge graph; setting up a consultation analysis directory; using a joint analysis of path blocking degree and semantic tension in the consultation analysis directory, combined with the association strength of obstacle nodes in the knowledge graph, to set up an initial progressive guidance tree structure; dynamically optimizing the initial progressive guidance tree structure by filtering entities and entity relationships to obtain a progressive guidance tree structure; and determining the consultation guidance strategy for the consultation scenario based on the semantic association paths and node association rules under the progressive guidance tree structure.
[0007] Preferably, a semantic association path is set based on the client's consultation tendency pointer and core consultation objective regarding the consultation content information; wherein, the path blockage degree is determined according to the number of obstacle nodes and the association strength in the semantic association path, and the semantic tension is quantified according to the semantic distance between the consultation tendency pointer and the core consultation objective.
[0008] Preferably, newly added entities and their relationships during multi-round consultation dialogues are captured, mapped to the knowledge graph, and the corresponding node association attributes are updated. If the newly added entity has a direct relationship with the core consultation objective, it is inserted as a new branch node into the corresponding level of the initial progressive guide tree structure, and the path blocking degree corresponding to the new branch node is adjusted according to the relationship strength.
[0009] Preferably, if the newly added entity forms a complementary relationship with the obstacle node, the association strength of the obstacle node is re-determined, and the dominant weight and priority ranking in the initial progressive guide tree structure are weakened; wherein, the basis for determining the complementary relationship is that the newly added entity and the obstacle node have compensatory semantic edges in the knowledge graph.
[0010] Preferably, when the newly added entity relationship constitutes an alternative semantic path to bypass the obstacle node, the bypass guidance tree structure is combined with the initial progressive guidance tree structure based on the semantic connectivity of each alternative semantic path in the knowledge graph and the generated bypass guidance tree structure.
[0011] Preferably, when combining the two branches, the effective branches in the bypass guidance tree structure and the initial progressive guidance tree structure whose path blocking degree is lower than the blocking degree threshold are retained, and the consultation guidance logic is dynamically expanded and optimized according to the shortest distance between the semantically related path and the core consultation objective.
[0012] Preferably, after the dynamic expansion and optimization of the consultation guidance logic, the semantic rationality of the entity relationship is verified: if the newly added entity relationship conflicts with the entity relationship corresponding to the initial progressive guidance tree structure, the relationship is corrected in combination with the semantic direction of the consultation tendency pointer, and the progressive guidance tree structure is obtained based on the corrected branch extension direction.
[0013] Preferably, a sequence of guiding statements is generated based on the hierarchical depth of the progressive guiding tree structure; the sequence of guiding statements is embedded in entity association examples in the knowledge graph; when it deviates from the core consultation objective, the cosine distance corresponding to the dialogue embedding vector under this round of consultation is analyzed; if it exceeds a preset offset threshold, a branch regression reminder is automatically triggered.
[0014] In a second aspect, this application provides a dynamic progressive consultation guidance system based on a knowledge graph, wherein the system comprises: an entity relationship extraction module: marking entities in a consultation scenario with multi-round consultation dialogues, extracting entity relationships from consultation content information corresponding to the multi-round consultation dialogues, and setting the knowledge graph according to the entities and entity relationships in the consultation scenario; a consultation analysis directory setting module: performing goal achievement assessment and obstacle node identification in the knowledge graph, and setting a consultation analysis directory; an association strength analysis module: using path blocking degree and semantic tension joint analysis in the consultation analysis directory, combined with the association strength of obstacle nodes in the knowledge graph, to set an initial progressive guidance tree structure; a dynamic optimization module: based on the initial progressive guidance tree structure, performing dynamic optimization by filtering entities and entity relationships to obtain a progressive guidance tree structure; and a consultation guidance strategy determination module: determining the consultation guidance strategy in the consultation scenario based on the semantic association paths and node association rules under the progressive guidance tree structure.
[0015] In summary, one or more technical solutions provided in this application achieve the technical effect of deeply analyzing and guiding the semantic association of consultation content by marking entities and extracting entity relationships from multi-round consultation dialogues, thereby improving the accuracy of identifying users' core consultation needs and making consultation guidance more targeted and flexible. Attached Figure Description
[0016] Figure 1 This application provides a flowchart illustrating a dynamic, progressive consultation guidance method based on knowledge graphs.
[0017] Figure 2 This application provides a schematic diagram of the structure of a dynamic progressive consultation guidance system based on knowledge graphs.
[0018] Explanation of reference numerals in the attached diagram: Entity Relationship Extraction Module M100, Consultation Analysis Catalog Setting Module M200, Association Strength Analysis Module M300, Dynamic Optimization Module M400, Consultation Guidance Strategy Determination Module M500. Detailed Implementation
[0019] Example 1: The present application will be described in detail below with reference to the accompanying drawings, as follows... Figure 1 As shown, this application provides a dynamic progressive consultation guidance method based on knowledge graphs, wherein the method includes: S1: Mark entities in the consultation scenario with multi-round consultation dialogues, extract entity relationships with consultation content information corresponding to the multi-round consultation dialogues, and set up the knowledge graph according to the entities and entity relationships in the consultation scenario; S2: Perform goal achievement evaluation and obstacle node identification in the knowledge graph, and set up a consultation analysis directory; S3: In the consultation analysis directory, use path blocking degree and semantic tension joint analysis, combined with the association strength of obstacle nodes in the knowledge graph, to set up an initial progressive guidance tree structure.
[0020] Specifically, in a consultation scenario, an entity refers to a specific object or concept directly related to the consultation content, such as product name, service type, or key terms in the user's question. For example, in a consultation scenario about electronic products, mobile phones, battery life, and screen resolution can all be considered entities. Entity relationships refer to the logical or semantic connections between entities. For example, there is a linear relationship between mobile phones and battery life, indicating that mobile phones have the attribute of battery life; there is a matching relationship between mobile phones and chargers, indicating that the charger is a matching device for the mobile phone. Path obstruction degree refers to the number and intensity of obstacle nodes on the path from one node to another in a knowledge graph. Obstacle nodes are those nodes that may prevent users from reaching their consultation goals, such as a user's misunderstanding of a concept or a specific question about a product. The higher the path obstruction degree, the greater the difficulty for the user to reach their consultation goals.
[0021] Semantic tension refers to the semantic distance between the user's consultation inclination and the core consultation goal in the consultation content. The smaller the semantic distance, the closer the user's consultation inclination is to the core consultation goal, and the smaller the semantic tension; conversely, the greater the semantic tension, the larger the semantic tension. The initial progressive guidance tree structure is a tree structure built based on knowledge graphs to represent the logical hierarchy and guidance path of the consultation content. Each node of the tree represents an entity or entity relationship, and the branches of the tree represent different consultation paths. The initial progressive guidance tree structure is constructed based on the results of joint analysis of path blocking degree and semantic tension, and is used to initially guide the user's consultation.
[0022] A knowledge graph is a structured semantic knowledge base that represents entities and the relationships between them in the form of a graph. In a consultation scenario, a knowledge graph is used to store and manage entities and their relationships within the consultation content, providing data support for consultation analysis. Goal achievement assessment refers to the quantitative evaluation of the degree to which a user has achieved their consultation goal. By analyzing the semantic relationship between the user's consultation content and the consultation goal, it assesses whether the user is close to or has achieved the consultation goal. Specifically, if a user is consulting about improving the battery life of their mobile phone, by analyzing the information in the user's dialogue, it assesses whether the user has obtained enough information to solve the problem. Obstacle node identification refers to identifying nodes in the knowledge graph that may hinder the user from achieving their consultation goal. Obstacle nodes are misunderstandings of a concept, confusion about a problem, or other obstacles in the consultation path. The consultation analysis catalog is an analytical framework set in the knowledge graph, used to organize and manage the consultation analysis process, including analytical tasks such as goal achievement assessment and obstacle node identification.
[0023] Execution steps: Using natural language processing (NLP) technology, key information is extracted from multi-round consultation dialogues, and entities in the consultation scenario are identified. Specifically, named entity recognition (NID) technology can be used to identify product names, key nouns in user questions, etc. The relationships between these entities are analyzed using techniques such as dependency parsing to determine logical or semantic connections. In the consultation analysis catalog, a joint analysis of path blocking degree and semantic tension is performed on each consultation path. Path blocking degree is determined by calculating the number of obstacle nodes and their association strength on the path; semantic tension is determined by quantifying the semantic distance between the user's consultation tendency and the core consultation objective.
[0024] Based on the analysis results of path obstruction and semantic tension, and combined with the association strength of obstacle nodes in the knowledge graph, an initial progressive guidance tree structure is constructed. The root node of the tree is typically the core consultation objective, while branch nodes are entities or entity relationships related to the core objective. The tree structure reflects the logical hierarchy and guidance sequence of the consultation path. In the above steps, by labeling entities and extracting entity relationships, the semantic associations of the consultation content can be deeply analyzed, providing accurate data support for subsequent analysis and guidance. The joint analysis of path obstruction and semantic tension can accurately locate obstacle nodes in the consultation and the user's core needs, thereby constructing an initial progressive guidance tree structure that can effectively guide user consultations, understand the user's consultation intent, provide a clear consultation path, improve consultation efficiency and user experience, and thus achieve accurate and efficient consultation guidance.
[0025] Entities and their relationships are extracted from the consultation scenario and stored in a knowledge graph. Natural language processing (NLP) technology is used to identify key entities and their relationships in the user's consultation. For example, key entity 1 is the mobile phone, key entity 2 is battery life, and the relationship between key entities 1 and 2 is the mobile phone's battery life. Based on the extracted entities and relationships, a graph structure of the knowledge graph is constructed, where each entity is a node and the relationships between entities are edges. By analyzing the semantic association between the user's consultation content and the consultation goal, the user's approximation or achievement of the consultation goal is assessed. Specifically, semantic similarity calculation methods are used to compare the semantic similarity between the user's current consultation content and the consultation goal, thereby evaluating the degree of goal achievement.
[0026] In a knowledge graph, by analyzing the semantic relationships between user consultation content and nodes in the knowledge graph, nodes that may hinder users from achieving their consultation goals are identified. For example, if a user misunderstands the concept of battery life, this misunderstanding can be identified as an obstacle node by analyzing keywords and semantic structure in the user's dialogue. Based on the results of goal achievement assessment and obstacle node identification, a consultation analysis catalog is set up. The catalog includes nodes and paths that require further analysis, as well as corresponding analysis methods and tools. For example, if a user's misunderstanding of the concept of battery life is identified, the consultation analysis catalog can include further explanations and guidance paths for the concept of battery life. In the above steps, goal achievement assessment and obstacle node identification help the system accurately locate users' consultation needs and potential problems, thereby providing users with more targeted and effective consultation guidance. Specifically, it quickly identifies the entities that users care about and their relationships, and assesses whether users are close to or have achieved their consultation goals, thereby improving consultation efficiency and user experience.
[0027] S4: Based on the initial progressive guidance tree structure, the entity and entity relationship are filtered and dynamically optimized to obtain the progressive guidance tree structure; S5: Based on the semantic association path and node association rules under the progressive guidance tree structure, the consultation guidance strategy in the consultation scenario is determined.
[0028] Specifically, dynamic optimization aims to optimize the initial progressive guidance tree structure based on real-time consultation dialogue content. By filtering and adjusting entities and their relationships, it ensures that the guidance tree structure can reflect new information and changes in user needs in real time. The progressive guidance tree structure is a dynamically optimized guidance tree structure that includes the initial guidance logic and can be dynamically adjusted according to the progress of the dialogue to better adapt to the user's consultation needs. The progressive guidance tree structure is more flexible and accurate, and can respond to changes in the dialogue in real time. Semantic association paths refer to the paths from one node to another in the knowledge graph, representing the semantic relationships between entities. They reflect the logic and semantic structure of the user's consultation content and are an important basis for determining consultation guidance strategies. Node association rules define the association methods and strengths between nodes in the knowledge graph, including the relationship types between entities, semantic distance, etc. Node association rules are used to guide how to construct and optimize the progressive guidance tree structure and how to generate guidance strategies based on the user's consultation content.
[0029] In the above steps, by dynamically optimizing the initial progressive guidance tree structure, it is possible to respond in real time to changes in user consultation content, ensuring that the guidance tree structure always reflects the latest dialogue information and user needs. The dynamic optimization process can handle newly added entities and entity relationships, and generate more accurate and flexible consultation guidance strategies based on semantic association paths and node association rules, thereby improving the targeting and flexibility of consultation guidance and effectively enhancing the accuracy of demand analysis and guidance efficiency in complex consultation scenarios.
[0030] Furthermore, in the consultation analysis catalog, a joint analysis of path blocking degree and semantic tension is adopted. The method of this application includes: Based on the client's consultation tendency pointer and core consultation objective, a semantic association path is set; wherein, the path blockage degree is determined according to the number of obstacle nodes and the association strength in the semantic association path, and the semantic tension is quantified according to the semantic distance between the consultation tendency pointer and the core consultation objective.
[0031] Specifically, the consultation tendency indicator refers to the tendency or focus shown by the client during the consultation process, reflecting the main direction and interest of the user's current consultation. For example, in the scenario of consulting about electronic products, if the user mentions battery life multiple times, then battery life can be regarded as the user's consultation tendency indicator. Semantic association path refers to the path from one node to another in the knowledge graph, representing the semantic association between these nodes, reflecting the logic and semantic structure of the user's consultation content.
[0032] Execution steps: Analyze user inquiries using natural language processing (NLP) technology to identify current inquiries. Specifically, use text classification or keyword extraction techniques to determine the entities or concepts the user is most interested in. If the user mentions battery life multiple times in the conversation, battery life can be identified as an indicator of inquiries. In the knowledge graph, starting from the inquiries indicator, find the path leading to the core inquiry goal. The nodes and edges on this path represent the semantic relationships of the user's inquiries. Specifically, if the core inquiry goal is how to improve the phone's battery life, and the inquiries indicator is battery life, the corresponding semantic relationship paths include paths under nodes such as charging habits and usage environment.
[0033] Path congestion is a comprehensive measure of the number of obstructive nodes and their association strength in a semantically related path. Obstructive nodes are those nodes that prevent users from reaching their consultation goals. Specifically, if a user has a misunderstanding about the concept of battery life, this misunderstanding is an obstructive node. Path congestion can be determined by calculating the number of obstructive nodes on the path and the association strength of each obstructive node. Association strength can be quantified by the semantic similarity between nodes or the user's attention to nodes. Semantic tension refers to the semantic distance between the consultation tendency indicator and the core consultation goal. The smaller the semantic distance, the closer the user's consultation tendency is to the core consultation goal, and the smaller the semantic tension; conversely, the greater the semantic tension, the larger the semantic tension. Furthermore, if a user is consulting about how to improve the battery life of a mobile phone, and the core consultation goal is the battery life of the mobile phone, then the semantic tension can be quantified by calculating the semantic similarity between the user's consultation content and the core consultation goal.
[0034] In the above steps, by extracting consultation tendency indicators and semantic connection paths, the system accurately understands the user's current consultation direction and focus, and assesses the semantic distance between the user and the consultation goal. The quantification of path obstruction degree and semantic tension further helps the system identify obstacles that the user may encounter during the consultation process, thereby providing data support for subsequent guidance strategies. Specifically, it identifies the consultation tendency indicators that the user is currently most concerned about and finds the semantic connection paths leading to the core consultation goal. If there are obstacle nodes in the path, the system can calculate the path obstruction degree and semantic tension to assess the impact of these obstacles on the user's consultation, and formulate more accurate and effective guidance strategies accordingly to help the user overcome obstacles and efficiently achieve the consultation goal.
[0035] Furthermore, based on the initial progressive guide tree structure, and using entities and entity relationships for dynamic optimization through filtering, the method of this application includes: Capture newly added entities and their relationships during multi-round consultation dialogues, map them to the knowledge graph, and update the corresponding node association attributes. If the newly added entity has a direct relationship with the core consultation objective, it is inserted as a new branch node into the corresponding level of the initial progressive guide tree structure. At the same time, the path blocking degree corresponding to the new branch node is adjusted according to the relationship strength.
[0036] Specifically, new entities refer to new concepts or objects that users may introduce as the conversation progresses in multiple rounds of consultation. These new concepts or objects are called new entities. For example, when inquiring about mobile phone battery life, a user might mention a charger brand, which is a new entity. New entity relationships refer to the new relationships that form between new entities and between new entities and existing entities. For example, there might be a compatibility relationship between the new entity "charger brand" and the existing entity "mobile phone". Node association attributes refer to the attributes that each node and edge in a knowledge graph has. These attributes describe the characteristics of nodes and edges, such as association strength and type. An entity node has an importance attribute, and a relationship edge has a semantic similarity attribute.
[0037] Execution steps: During multi-round consultation dialogues, the system continuously monitors the dialogue content and uses natural language processing technology to identify newly added entities and entity relationships. For example, when a user mentions that they recently switched to a charger from brand A, the system can identify brand A charger as a new entity and the relationship between this new entity and the phone's battery life. The new entities and relationships are mapped to a knowledge graph, and the structure and content of the knowledge graph are updated to ensure that the knowledge graph can reflect the latest information in the dialogue in real time. For newly added entities and relationships, initial association attributes, such as association strength, are assigned to them based on their frequency of appearance in the dialogue and the user's level of attention.
[0038] Simultaneously, the association attributes of existing nodes and relationships are adjusted based on the newly added entities and relationships. Furthermore, since the newly added entity, "Brand A charger," has a strong effect relationship with the existing entity, "Mobile phone battery life," the semantic similarity attribute value of the relationship between these two entities is enhanced. The system determines whether there is a direct association between the newly added entity and the core consultation objective. If so, the newly added entity, "Brand A charger," is directly associated with the core consultation objective of improving mobile phone battery life, and is then inserted as a new branch node into the corresponding level of the initial progressive guide tree structure. Based on the strength of the relationship between the newly added entity and the core consultation objective, the path blocking degree corresponding to the new branch node is adjusted. Furthermore, since the relationship strength between "Brand A charger" and improving mobile phone battery life is relatively high, indicating that the newly added entity is of great help in achieving the consultation objective, the path blocking degree corresponding to the new branch node is reduced, thereby highlighting the importance of this path in the guide tree structure.
[0039] In the above steps, by capturing new entities and relationships in multiple rounds of consultation dialogue and updating the knowledge graph and progressive guidance tree structure in a timely manner, the system responds to changes in user consultation content in real time, ensuring that the guidance strategy is always based on the latest information. This not only improves the accuracy of consultation guidance and enhances the system's flexibility and user experience, but also provides users with guidance suggestions that are more relevant to the current dialogue content, thereby more effectively helping users solve consultation problems and achieve consultation goals.
[0040] Furthermore, the method of this application includes: If the newly added entity forms a complementary relationship with the obstacle node, the association strength of the obstacle node is re-determined, and its dominant weight and priority ranking in the initial progressive guiding tree structure are weakened; wherein, the basis for determining the complementary relationship is that the newly added entity and the obstacle node have compensatory semantic edges in the knowledge graph.
[0041] Specifically, in a knowledge graph, a complementary relationship refers to a relationship between two entities that complement each other, indicating that one entity can provide information or functionality lacking in the other. For example, if an obstacle node represents a user's misunderstanding of battery life, and a new entity provides the correct explanation or relevant information, then there is a complementary relationship between the two entities. A compensatory semantic edge is an edge connecting two entities in a knowledge graph, indicating a complementary relationship. The existence of a compensatory semantic edge shows that one entity can compensate for the shortcomings of the other, thus helping to resolve obstacles in the user's inquiry. Association strength refers to the closeness of the relationship between two entities. In a knowledge graph, association strength can be quantified in various ways, such as by the frequency with which users mention the two entities simultaneously, semantic similarity, etc. Higher association strength indicates a closer relationship between the two entities. Dominant weight refers to the importance of a node or path in a progressive guidance tree structure. Nodes or paths with high dominant weights have higher priority in the guidance tree and have a greater influence on the direction of user inquiry guidance.
[0042] Execution steps: In the knowledge graph, check whether there is a compensatory semantic edge between the new entity and the obstacle node. Specifically, if the new entity provides a correct explanation of the concept of battery life, and the obstacle node represents a user's misunderstanding of the concept of battery life, then there is a compensatory semantic edge between the two entities. If a compensatory semantic edge exists, the system will identify this complementary relationship and label the new entity and the obstacle node. Based on the complementary relationship, the association strength of the obstacle node is re-evaluated. Furthermore, since the new entity provides compensatory information, the association strength of the obstacle node will decrease. For example, if the new entity provides a correct explanation, the association strength of the user's misunderstanding of the concept of battery life will decrease from a high value to a low value. The re-evaluation can be achieved by calculating factors such as the semantic similarity between the new entity and the obstacle node, and the user's attention to the new entity.
[0043] In the progressive guidance tree structure, the dominant weight and priority ranking of obstacle nodes are adjusted based on the redefined association strength. As the association strength of an obstacle node decreases, its dominant weight in the guidance tree also decreases accordingly, and its priority ranking shifts to the next level. In the above steps, by identifying the complementary relationships between new entities and obstacle nodes, and accordingly redetermining the association strength of obstacle nodes, reducing their association strength and dominant weight, the guidance path in the guidance tree is adjusted more rationally. This allows for more accurate assessment and handling of obstacles in user consultations, improving the accuracy of consultation guidance, efficiently achieving consultation goals, and enhancing consultation efficiency and user experience.
[0044] Furthermore, the method of this application includes: When the newly added entity relationship constitutes an alternative semantic path to bypass the obstacle node, the bypass guidance tree structure is combined with the initial progressive guidance tree structure based on the semantic connectivity of each alternative semantic path in the knowledge graph and the generated bypass guidance tree structure.
[0045] Specifically, in a knowledge graph, an alternative semantic path refers to a path from one node to another that bypasses existing obstacle nodes and provides a new semantic association. For instance, in an obstacle node, a user may misunderstand the concept of battery life, while a newly added entity relationship provides a new explanation; this new path is the alternative semantic path. Semantic connectivity refers to the degree of semantic association between different paths in a knowledge graph, measuring the logical and semantic coherence between paths. Furthermore, if two paths both involve battery life and charging habits but are connected through different intermediate nodes, the semantic connectivity between the two paths is high. A bypass guidance tree structure refers to a new guidance tree structure generated based on alternative semantic paths, in addition to the initial progressive guidance tree structure. The bypass guidance tree structure provides guidance paths that bypass obstacle nodes, helping users achieve their consultation goals more effectively.
[0046] Execution steps: In the knowledge graph, check whether the newly added entity relationship constitutes an alternative semantic path to bypass the obstacle node. Specifically, if the user has a misunderstanding of the concept of battery life, and the newly added entity relationship provides a new association between charging habits and battery life, this new path is an alternative semantic path. By analyzing the semantic connectivity in the knowledge graph, determine the effectiveness of these alternative semantic paths. Furthermore, if the newly added entity relationship provides a strong semantic association between charging habits and battery life, the effectiveness of the alternative semantic path is relatively high.
[0047] Based on the identified alternative semantic paths, a bypass guidance tree structure is generated. The bypass guidance tree structure includes newly added entities and the relationships between them, as well as how these relationships bypass obstacle nodes. The generated bypass guidance tree structure is combined with the initial progressive guidance tree structure. During the combination process, effective branches with path blocking degree below the blocking degree threshold are retained, and the consultation guidance logic is dynamically expanded and optimized according to the shortest distance between the semantically related path and the core consultation goal. Furthermore, if there is a path with high blocking degree in the initial progressive guidance tree structure, and the bypass guidance tree structure provides a path with low blocking degree, this new path will be integrated into the initial progressive guidance tree structure to optimize the overall guidance logic.
[0048] In the above steps, by identifying alternative semantic paths formed by newly added entity relationships and generating a bypass guidance tree structure, new guidance paths that bypass obstacle nodes can be provided, improving the accuracy and flexibility of consultation guidance, enhancing system adaptability, generating a bypass guidance tree structure, and integrating it into the initial progressive guidance tree structure, enabling the system to more effectively help users overcome obstacles, efficiently achieve consultation goals, and improve consultation efficiency and user experience.
[0049] Furthermore, by combining the bypass guidance tree structure with the initial progressive guidance tree structure, the method of this application includes: When combining the two branches, the effective branches with path blocking degree below the blocking degree threshold in the bypass guidance tree structure and the initial progressive guidance tree structure are retained, and the consultation guidance logic is dynamically expanded and optimized according to the shortest distance between the semantically related path and the core consultation objective.
[0050] Specifically, the blocking threshold is used to determine whether the path blocking degree is within an acceptable range. If the path blocking degree is below the blocking threshold, it indicates that the path is relatively unobstructed and can be retained as a valid path. Specifically, if the blocking threshold is set to 0.5, then paths with a blocking degree below 0.5 are considered valid paths. Semantic association paths refer to the paths from one node to another in the knowledge graph, representing the semantic relationships between these nodes. They reflect the logic and semantic structure of the user's consultation content and are an important basis for determining the consultation guidance strategy. Dynamic expansion and optimization of consultation guidance logic refers to dynamically adjusting and optimizing the consultation guidance logic based on real-time consultation dialogue content and knowledge graph updates. By selecting paths with low blocking degree and short semantic association paths, users can be guided to achieve their consultation goals more efficiently.
[0051] Execution steps: Evaluate each path in the initial progressive guidance tree structure and the bypass guidance tree structure, and calculate its path blocking degree. The path blocking degree reflects the number of obstacle nodes and the strength of their associations on the path. For example, if a path has multiple obstacle nodes and the association strength of these nodes is high, then the path blocking degree will be high. Retain valid branches with a path blocking degree below the blocking degree threshold. Among the retained valid branches, select the path with the shortest semantic association path. The length of the semantic association path can be determined by calculating the semantic distance between nodes on the path. Specifically, if the semantic similarity between nodes on a certain path is high, then the semantic association path of this path is short. Select the shortest semantic association path to ensure that users can achieve their consultation goals in the most direct way and reduce unnecessary detours and obstacles.
[0052] The selected shortest semantic association path is integrated into the initial progressive guidance tree structure to form a new progressive guidance tree structure, which more efficiently guides users to achieve their consultation goals. In the above steps, by evaluating the path obstruction degree and selecting the shortest semantic association path, it is integrated into the initial progressive guidance tree structure, dynamically expanding and optimizing the consultation guidance logic, ensuring the smooth and efficient guidance path, more effectively helping users overcome obstacles, efficiently achieve consultation goals, improving the accuracy and flexibility of consultation guidance, and also enhancing the system's adaptability and user experience.
[0053] Furthermore, to obtain a progressive guide tree structure, the method of this application includes: After the dynamic expansion and optimization of the consultation guidance logic, the semantic rationality of the entity relationship is verified: if the newly added entity relationship conflicts with the entity relationship corresponding to the initial progressive guidance tree structure, the relationship is corrected in combination with the semantic direction of the consultation tendency pointer, and the progressive guidance tree structure is obtained based on the corrected branch extension direction.
[0054] Specifically, semantic rationality refers to the logicality and consistency of entity relationships in terms of semantics, ensuring that the relationships in the knowledge graph are not only grammatically correct but also semantically consistent with common sense and logic. Relationship correction refers to adjusting and correcting these relationships when conflicts are found between newly added entity relationships and the entity relationships corresponding to the initial progressive guide tree structure. The purpose of correction is to ensure that the relationships in the knowledge graph are semantically rational and consistent. Branch extension direction refers to the insertion direction of new branch nodes in the progressive guide tree structure, which determines the position and level of the new branch nodes in the tree structure, thereby affecting the structure and logic of the progressive guide tree structure.
[0055] Execution steps: In the dynamically expanded and optimized consultation guidance logic, the semantic rationality of newly added entity relationships is checked. The purpose of the check is to ensure that the newly added entity relationships are semantically consistent with existing entity relationships and do not conflict. For example, if the newly added entity relationship indicates that there is a parallel relationship between charger brand and battery life, while the initial progressive guidance tree structure shows a linear relationship between charger brand and battery life, then these two relationships conflict. If a conflict is found between the newly added entity relationship and the entity relationship corresponding to the initial progressive guidance tree structure, the relationship will be corrected in conjunction with the semantic direction of the consultation tendency pointer to ensure semantic rationality.
[0056] After correcting the relationships, the new branch nodes are inserted into the corresponding levels of the progressive guided tree structure according to the corrected branch extension direction. Specifically, if the corrected relationship between the charger brand and battery life is linear, the charger brand will be inserted as a new branch node at the next level below the battery life node, thus forming a new progressive guided tree structure. In the above steps, semantic rationality verification and relationship correction ensure the semantic consistency and logic of the knowledge graph and the progressive guided tree structure, avoiding logical errors or conflicts in the knowledge graph, improving the quality and reliability of the knowledge graph, and enhancing the accuracy and effectiveness of consultation guidance.
[0057] Furthermore, based on the semantic association paths and node association rules under the aforementioned progressive guidance tree structure, the consultation guidance strategy for the consultation scenario is determined. The method of this application also includes: Based on the hierarchical depth of the progressive guidance tree structure, a sequence of guidance statements is generated. The sequence of guidance statements is embedded in entity association examples in the knowledge graph. When it deviates from the core consultation goal, the cosine distance corresponding to the dialogue embedding vector under this round of consultation is analyzed. If it exceeds a preset offset threshold, a branch regression reminder is automatically triggered.
[0058] Specifically, hierarchy depth refers to the path length from the root node corresponding to the core consultation goal to a certain node in the progressive guidance tree structure. It reflects the position and importance of the node in the tree structure. In a consultation scenario, the root node is improving mobile phone battery life, the first level is charging habits, and the second level is charger brands, etc.; guidance statement sequence refers to a series of guidance statements generated according to the progressive guidance tree structure, used to guide users to gradually achieve the consultation goal. Each statement corresponds to a node or path in the tree structure; entity association example refers to specific association instances between entities in the knowledge graph, used to enhance the understandability and practicality of guidance statements; dialogue embedding vector refers to converting dialogue content into a high-dimensional vector representation, used to quantify the semantic information of the dialogue content. Through dialogue embedding vector, the semantic similarity between dialogue content can be calculated; cosine distance refers to the cosine value of the angle between two vectors, used to measure the similarity between two vectors. The smaller the cosine distance, the more similar the two vectors are; the larger the cosine distance, the less similar the two vectors are; branch regression reminder refers to an automatic reminder mechanism that is triggered when the user's consultation content deviates from the core consultation goal, reminding the user to return to the guidance path more relevant to the core consultation goal.
[0059] Execution Steps: Based on the hierarchical depth of the progressive guidance tree structure, starting from the root node corresponding to the core consultation goal, guidance statements are generated layer by layer. Each level of guidance statement corresponds to a node or path in the tree structure. Entity association examples from the knowledge graph are embedded in the generated guidance statements to enhance their comprehensibility and usability. After each round of consultation dialogue, the user's dialogue content is converted into a dialogue embedding vector, and the cosine distance between this vector and the core consultation goal vector is obtained. If the calculated cosine distance exceeds a preset offset threshold, it indicates that the user's consultation content has deviated from the core consultation goal, and a branch regression reminder is automatically triggered. In the above steps, by generating a sequence of guidance statements and embedding entity association examples, a clear and specific guidance path is provided to the user, helping them gradually achieve their consultation goal. Simultaneously, by analyzing the dialogue embedding vector and triggering branch regression reminders, the system monitors in real time whether the user's consultation content deviates from the core goal, providing timely reminders when the user deviates, guiding the user back on track, ensuring that the user can efficiently achieve their consultation goal, and ensuring the accuracy and effectiveness of the consultation guidance.
[0060] In summary, the beneficial effects of the embodiments of this application are: This application provides a dynamic, progressive consultation guidance method and system based on a knowledge graph. It achieves the technical effect of marking entities in a consultation scenario through multi-turn consultation dialogues, extracting entity relationships from the consultation content information corresponding to these dialogues, and setting up a knowledge graph based on the entities and their relationships within the consultation scenario. Within the knowledge graph, it performs goal achievement assessment and obstacle node identification, and establishes a consultation analysis directory. In this directory, it employs joint analysis of path obstruction and semantic tension, combined with the association strength of obstacle nodes in the knowledge graph, to set an initial progressive guidance tree structure. Based on this initial progressive guidance tree structure, it dynamically optimizes the selection of entities and their relationships to obtain a further progressive guidance tree structure. Finally, it determines the consultation guidance strategy for the consultation scenario based on the semantic association paths and node association rules under the progressive guidance tree structure.
[0061] Example 2, based on the same inventive concept as the knowledge graph-based dynamic progressive consultation guidance method in the aforementioned examples, such as... Figure 2 As shown in the embodiment of this application, a dynamic progressive consultation guidance system based on knowledge graphs is provided, wherein the system includes: Entity Relationship Extraction Module M100: Marks entities in the consultation scenario with multi-round consultation dialogues, extracts entity relationships with consultation content information corresponding to the multi-round consultation dialogues, and sets up the knowledge graph based on the entities and entity relationships in the consultation scenario.
[0062] Consultation Analysis Catalog Setting Module M200: Performs goal achievement assessment and obstacle node identification in the knowledge graph, and sets the consultation analysis catalog.
[0063] Association Strength Analysis Module M300: In the consultation analysis directory, a joint analysis of path blocking degree and semantic tension is adopted, and the association strength of obstacle nodes in the knowledge graph is combined to set an initial progressive guide tree structure.
[0064] Dynamic optimization module M400: Based on the initial progressive guide tree structure, it performs dynamic optimization by filtering entities and entity relationships to obtain a progressive guide tree structure.
[0065] Consultation guidance strategy determination module M500: Determines the consultation guidance strategy in the consultation scenario based on the semantic association path and node association rules under the progressive guidance tree structure.
[0066] Furthermore, the correlation strength analysis module M300 is also used to perform the following method: Based on the client's consultation tendency pointer and core consultation objective, a semantic association path is set; wherein, the path blockage degree is determined according to the number of obstacle nodes and the association strength in the semantic association path, and the semantic tension is quantified according to the semantic distance between the consultation tendency pointer and the core consultation objective.
[0067] Furthermore, the dynamic optimization module M400 is used to perform the following method: Capture newly added entities and their relationships during multi-round consultation dialogues, map them to the knowledge graph, and update the corresponding node association attributes. If the newly added entity has a direct relationship with the core consultation objective, it is inserted as a new branch node into the corresponding level of the initial progressive guide tree structure. At the same time, the path blocking degree corresponding to the new branch node is adjusted according to the relationship strength.
[0068] Furthermore, the dynamic optimization module M400 is also used to perform the following methods: If the newly added entity forms a complementary relationship with the obstacle node, the association strength of the obstacle node is re-determined, and its dominant weight and priority ranking in the initial progressive guiding tree structure are weakened; wherein, the basis for determining the complementary relationship is that the newly added entity and the obstacle node have compensatory semantic edges in the knowledge graph.
[0069] Furthermore, the dynamic optimization module M400 is also used to perform the following methods: When the newly added entity relationship constitutes an alternative semantic path to bypass the obstacle node, the bypass guidance tree structure is combined with the initial progressive guidance tree structure based on the semantic connectivity of each alternative semantic path in the knowledge graph and the generated bypass guidance tree structure.
[0070] Furthermore, the dynamic optimization module M400 is also used to perform the following methods: When combining the two branches, the effective branches with path blocking degree below the blocking degree threshold in the bypass guidance tree structure and the initial progressive guidance tree structure are retained, and the consultation guidance logic is dynamically expanded and optimized according to the shortest distance between the semantically related path and the core consultation objective.
[0071] Furthermore, the dynamic optimization module M400 is also used to perform the following methods: After the dynamic expansion and optimization of the consultation guidance logic, the semantic rationality of the entity relationship is verified: if the newly added entity relationship conflicts with the entity relationship corresponding to the initial progressive guidance tree structure, the relationship is corrected in combination with the semantic direction of the consultation tendency pointer, and the progressive guidance tree structure is obtained based on the corrected branch extension direction.
[0072] Furthermore, the consultation guidance strategy determination module M500 is also used to perform the following methods: Based on the hierarchical depth of the progressive guidance tree structure, a sequence of guidance statements is generated. The sequence of guidance statements is embedded in entity association examples in the knowledge graph. When it deviates from the core consultation goal, the cosine distance corresponding to the dialogue embedding vector under this round of consultation is analyzed. If it exceeds a preset offset threshold, a branch regression reminder is automatically triggered.
[0073] In summary, any step can be stored as a computer instruction or program in an unrestricted computer memory and can be called and recognized by an unrestricted computer processor; no further restrictions are imposed here.
[0074] Furthermore, the above technical solutions only embody the preferred technical solutions of the embodiments of this application. Any changes that those skilled in the art may make to certain parts of these solutions embody the novel principles of the embodiments of this application. Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application.
Claims
1. A dynamic, progressive consultation guidance method based on knowledge graphs, characterized in that: The method includes: Entities in a consultation scenario are marked by multi-round consultation dialogues, entity relationships are extracted from the consultation content information corresponding to the multi-round consultation dialogues, and the knowledge graph is set according to the entities and entity relationships in the consultation scenario. The goal achievement rate is assessed and obstacle nodes are identified within the knowledge graph, and a consultation analysis catalog is set up. In the consultation analysis catalog, a joint analysis of path blocking degree and semantic tension is adopted, and the initial progressive guidance tree structure is set by combining the association strength of obstacle nodes in the knowledge graph. Based on the initial progressive guide tree structure, the progressive guide tree structure is obtained by dynamically optimizing the filtering of entities and entity relationships. Based on the semantic association paths and node association rules under the progressive guidance tree structure, the consultation guidance strategy in the consultation scenario is determined.
2. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 1, characterized in that, The consultation analysis catalog employs a joint analysis of path blocking degree and semantic tension, the method of which includes: Based on the client's consultation preference indicators and core consultation objectives regarding the consultation content information, a semantic association path is set; The path obstruction degree is determined by the number and association strength of the obstacle nodes in the semantic association path, and the semantic tension is quantified by the semantic distance between the consultation tendency pointer and the core consultation goal.
3. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 2, characterized in that, Based on the initial progressive guided tree structure, the method involves dynamic optimization through filtering of entities and entity relationships, including: Capture newly added entities and relationships in multi-round consultation dialogues, map them to the knowledge graph, and update the corresponding node association attributes; If the newly added entity has a direct relationship with the core objective of the consultation, it will be inserted as a new branch node into the corresponding level of the initial progressive guide tree structure, and the path blocking degree corresponding to the new branch node will be adjusted according to the strength of the relationship.
4. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 3, characterized in that, The method includes: If the newly added entity forms a complementary relationship with the obstacle node, the association strength of the obstacle node is re-determined, and the dominant weight and priority ranking in the initial progressive guide tree structure are weakened. The basis for determining the complementary relationship is that the newly added entity and the obstacle node have compensatory semantic edges in the knowledge graph.
5. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 4, characterized in that, The method includes: When the newly added entity relationship constitutes an alternative semantic path to bypass the obstacle node, the bypass guidance tree structure is combined with the initial progressive guidance tree structure based on the semantic connectivity of each alternative semantic path in the knowledge graph and the generated bypass guidance tree structure.
6. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 5, characterized in that, The method of combining the bypass guidance tree structure with the initial progressive guidance tree structure includes: When combining the two branches, the effective branches with path blocking degree below the blocking degree threshold in the bypass guidance tree structure and the initial progressive guidance tree structure are retained, and the consultation guidance logic is dynamically expanded and optimized according to the shortest distance between the semantically related path and the core consultation objective.
7. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 6, characterized in that, The method for obtaining a progressive guide tree structure includes: After the dynamic expansion and optimization of the consultation guidance logic, the semantic rationality of the entity relationship is verified: if the newly added entity relationship conflicts with the entity relationship corresponding to the initial progressive guidance tree structure, the relationship is corrected in combination with the semantic direction of the consultation tendency pointer, and the progressive guidance tree structure is obtained based on the corrected branch extension direction.
8. The dynamic progressive consultation guidance method based on knowledge graphs as described in claim 7, characterized in that, Based on the semantic association paths and node association rules under the progressive guidance tree structure, the consultation guidance strategy in the consultation scenario is determined, and the method further includes: Generate a sequence of guiding statements based on the hierarchical depth of the progressive guiding tree structure; The example of embedding the guiding statement sequence into the entity association in the knowledge graph, when it deviates from the core consultation goal, analyzes the cosine distance corresponding to the dialogue embedding vector under this round of consultation dialogue. If it exceeds the preset offset threshold, a branch regression reminder is automatically triggered.
9. A dynamic progressive consultation guidance system based on knowledge graphs, characterized in that: The system is used to implement the knowledge graph-based dynamic progressive consultation guidance method according to any one of claims 1-8, wherein the system comprises: Entity Relationship Extraction Module: Marks entities in the consultation scenario with multi-round consultation dialogues, extracts entity relationships with consultation content information corresponding to the multi-round consultation dialogues, and sets the knowledge graph according to the entities and entity relationships in the consultation scenario; Consultation Analysis Catalog Setting Module: This module assesses goal achievement and identifies obstacle nodes within the knowledge graph, and sets up the consultation analysis catalog. Association Strength Analysis Module: In the consultation analysis directory, a joint analysis of path blocking degree and semantic tension is adopted, and the association strength of obstacle nodes in the knowledge graph is combined to set an initial progressive guidance tree structure; Dynamic optimization module: Based on the initial progressive guide tree structure, the module performs dynamic optimization by filtering entities and entity relationships to obtain the progressive guide tree structure. Consultation guidance strategy determination module: Based on the semantic association path and node association rules under the progressive guidance tree structure, determine the consultation guidance strategy in the consultation scenario.