Question and answer method and system based on multi-view knowledge graph

Abstract

The invention discloses a question answering method and system based on a multi-view knowledge graph, and the method comprises the steps: firstly constructing the multi-view knowledge graph, generating knowledge graphs under different levels of view, and then automatically evaluating the level, namely the knowledge level, of a user in an interactive conversation process with the user. And calling the knowledge graph under the corresponding level perspective to search the answer, so that on the premise of not influencing the dialogue quality, the operation resources can be effectively saved, the response speed is improved, the flexibility and adaptability of the question answering system are improved, the intelligence of the question answering system is enhanced, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and specifically to a question-answering method and system based on a multi-perspective knowledge graph. Background Art

[0002] At present, with the development of artificial intelligence, the automatic question-answering system has gradually become a new mode of communication between people and machines. After understanding the questions input by the user, it can search for answers to the user's questions in the knowledge graph and return accurate answers to the questions.

[0003] Knowledge graph-based question-answering systems usually require searching the knowledge graph. However, when the knowledge graph becomes increasingly large, the time required for searching and filtering is difficult to meet real-time requirements. At the same time, when facing user groups with different knowledge backgrounds, the amount of basic knowledge required to support question-answering also varies. In this case, the strategy of relying solely on a single knowledge graph to support question-answering is obviously not flexible enough. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a question-answering method and system based on a multi-perspective knowledge graph, construct knowledge graphs under different levels of perspective, and select the knowledge graph under the corresponding level of perspective for answer search based on the judgment of the user level, which effectively saves the computing resources of the question-answering system and improves the response speed of the question-answering system.

[0005] The technical solution of the present invention is:

[0006] A question-answering method based on a multi-view knowledge graph specifically includes the following steps:

[0007] S1. Construct a multi-perspective knowledge graph, where G = {V, E}, where V represents a node in the multi-perspective knowledge graph. E represents the edge connecting the nodes, and R represents the relationship between the nodes. All nodes in the original knowledge graph are divided into different levels, and knowledge graphs at different levels are generated;

[0008] S2, conduct multiple rounds of interactive dialogues with the user, record the levels corresponding to the nodes involved in each round of dialogue, and calculate the user level based on the distribution of nodes of each level in the interactive dialogue with the user;

[0009] S3. Based on the calculated user level, select the knowledge graph under the level perspective corresponding to the user level, and then search in the knowledge graph under the level perspective to obtain the answer to the question raised by the user.

[0010] The specific process of dividing all nodes into different levels and generating knowledge graphs from different level perspectives is as follows: iFrom the perspective of the original knowledge graph, the nodes in the original knowledge graph are divided into levels not higher than level L. i and above level L i Two classes, then cleared above grade L i The node N of the original knowledge graph is connected to other nodes adjacent to any node N, and a newly added edge is obtained. The newly added edge is the edge directly connecting any node N with other adjacent nodes in the original knowledge graph. The level of other adjacent nodes is not higher than level L. i Repeat the above steps until all nodes above level L are i Clear all nodes N with a level no higher than level L in the original knowledge graph. i The nodes and the newly added edges together constitute the level L i Knowledge graph from perspective.

[0011] The level corresponding to the nodes involved in each round of dialogue is to extract entities from each round of dialogue, and then search for the entity in the original knowledge graph to obtain the level of the node corresponding to the entity.

[0012] The specific calculation method of the user level is shown in the following formula (1):

[0013]

[0014] In formula (1), L is the user level, m represents the number of nodes that appear in the interactive dialogue with the user, c i represents the number of times node i appears in the conversation, d i represents the degree of node i in the original knowledge graph, L i Indicates the level of node i.

[0015] In the interactive dialogue with the user, a user feedback mechanism is introduced, and the multi-perspective knowledge graph is continuously optimized through user evaluation and feedback, thereby improving the performance of the question-answering system.

[0016] A question-answering system based on a multi-view knowledge graph, comprising a multi-view knowledge graph construction module, an interactive receiving module, a node parsing and grading module, a user level calculation module, an extraction and return module, and a user feedback module;

[0017] The multi-view knowledge graph construction module is used to construct and optimize the multi-view knowledge graph;

[0018] The interactive receiving module is used to conduct interactive multi-round dialogues with users, receive questions from users, and receive user evaluations and feedback;

[0019] The node parsing and grading module is used to parse and obtain the corresponding grades of the nodes involved in each round of dialogue;

[0020] The user level calculation module is used to calculate the user level according to formula (1);

[0021] The extraction and return module is used to search and obtain answers to questions raised by users in the knowledge graph from the perspective of the corresponding level;

[0022] The user feedback module is used to obtain user evaluations and feedback, and continuously optimize the multi-view knowledge graph.

[0023] Advantages of the present invention:

[0024] (1) During the interactive dialogue with the user, the present invention automatically evaluates the user's level, that is, the knowledge level, and calls the knowledge graph from the corresponding level perspective to search for answers. This can effectively save computing resources and improve response speed without affecting the quality of the dialogue. This not only improves the flexibility and adaptability of the question-answering system, but also enhances the intelligence of the question-answering system and the user experience.

[0025] (2) The multi-perspective knowledge graph included in the present invention will be continuously optimized based on user evaluation and feedback, so that the question-answering system has stronger understanding and generation capabilities, and can provide rich and relevant knowledge information during the conversation process, further improving user satisfaction and system performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of the question-answering method of the present invention.

[0027] Figure 2 It is a schematic diagram of the logical structure of the question-answering system of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] See Figure 1 , a question answering method based on a multi-view knowledge graph, specifically comprising the following steps:

[0030] S1. Construct a multi-perspective knowledge graph, where G = {V, E}, where V represents a node in the multi-perspective knowledge graph. E represents the edge connecting the nodes, and R represents the relationship between the nodes. All nodes in the original knowledge graph are divided into different levels, and knowledge graphs from different levels are generated. The specific process is as follows:

[0031] At level L i From the perspective of the original knowledge graph, the nodes in the original knowledge graph are divided into levels not higher than level L. i and above level L i Two classes, then cleared above grade L i At the same time, the nodes adjacent to any node N are connected to obtain the newly added edges, where Li represents any level of the node in the original knowledge graph. The newly added edges are the edges directly connecting any node N with other adjacent nodes in the original knowledge graph, and the adjacent nodes are all at levels not higher than level L. i Repeat the above steps until all nodes above level L are i Clear all nodes N with a level no higher than level L in the original knowledge graph. i The nodes and the newly added edges together constitute the level L i Knowledge graph from perspective;

[0032] S2. Conduct multiple rounds of interactive dialogues with the user, extract entities from each round of dialogue, search for the entity in the original knowledge graph to obtain the level of the node corresponding to the entity, and calculate the user level based on the distribution of nodes of each level in the interactive dialogue with the user. The specific calculation method is shown in the following formula (1):

[0033]

[0034] In formula (1), L is the user level, m represents the number of nodes that appear in the interactive dialogue with the user, c i represents the number of times node i appears in the conversation, d i represents the degree of node i in the original knowledge graph, L i Indicates the level of node i;

[0035] S3. Based on the calculated user level, select the knowledge graph under the level perspective corresponding to the user level, and then search in the knowledge graph under the level perspective to obtain the answer to the question raised by the user.

[0036] Among them, a user feedback mechanism is introduced in the interactive dialogue with users, and the multi-perspective knowledge graph is continuously optimized through user evaluation and feedback, thereby improving the understanding and generation capabilities of the question-answering system. That is, the levels of nodes are re-divided according to user feedback, and knowledge graphs at different levels of perspective are generated.

[0037] See Figure 2 , a question-answering system based on a multi-view knowledge graph, comprising a multi-view knowledge graph construction module 1, an interaction receiving module 2, a node parsing and grading module 3, a user level calculation module 4, an extraction and return module 5 and a user feedback module 6;

[0038] The multi-view knowledge graph construction module 1 is used to construct and optimize the multi-view knowledge graph;

[0039] The interactive receiving module 2 is used to conduct interactive multi-round dialogues with users, receive questions from users, and receive user evaluations and feedback;

[0040] The node analysis and grading module 3 is used to analyze and obtain the corresponding grades of the nodes involved in each round of dialogue;

[0041] The user level calculation module 4 is used to calculate the user level according to formula (1);

[0042] The extraction and return module 5 is used to search and obtain the answer to the question raised by the user in the knowledge graph under the corresponding level perspective;

[0043] The user feedback module 6 is used to obtain user evaluations and feedback, and continuously optimize the multi-view knowledge graph.

[0044] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A question-answering method based on a multi-view knowledge graph, characterized in that: The specific steps include: S1. Construct a multi-perspective knowledge graph, where G = {V, E}, where V represents a node in the multi-perspective knowledge graph. E represents the edge connecting the nodes, and R represents the relationship between the nodes. All nodes in the original knowledge graph are divided into different levels, and knowledge graphs at different levels are generated; S2, conduct multiple rounds of interactive dialogues with the user, record the levels corresponding to the nodes involved in each round of dialogue, and calculate the user level based on the distribution of nodes of each level in the interactive dialogue with the user; S3. Based on the calculated user level, select the knowledge graph under the level perspective corresponding to the user level, and then search in the knowledge graph under the level perspective to obtain the answer to the question raised by the user.

2. A question-answering method based on a multi-view knowledge graph according to claim 1, characterized in that: The specific process of dividing all nodes into different levels and generating knowledge graphs from different level perspectives is as follows: i From the perspective of the original knowledge graph, the nodes in the original knowledge graph are divided into levels not higher than level L. i and above level L i Two classes, then cleared above grade L i The node N of the original knowledge graph is connected to other nodes adjacent to any node N, and a newly added edge is obtained. The newly added edge is the edge directly connecting any node N with other adjacent nodes in the original knowledge graph. The level of other adjacent nodes is not higher than level L. i Repeat the above steps until all nodes above level L are i Clear all nodes N with a level no higher than level L in the original knowledge graph. i The nodes and the newly added edges together constitute the level L i Knowledge graph from perspective.

3. A question-answering method based on a multi-view knowledge graph according to claim 2, characterized in that: The level corresponding to the nodes involved in each round of dialogue is to extract entities from each round of dialogue, and then search for the entity in the original knowledge graph to obtain the level of the node corresponding to the entity.

4. A question-answering method based on a multi-view knowledge graph according to claim 3, characterized in that: The specific calculation method of the user level is shown in the following formula (1): In formula (1), L is the user level, m represents the number of nodes that appear in the interactive dialogue with the user, c i represents the number of times node i appears in the conversation, d i represents the degree of node i in the original knowledge graph, L i Indicates the level of node i.

5. The question-answering method based on a multi-view knowledge graph according to claim 1, characterized in that: In the interactive dialogue with the user, a user feedback mechanism is introduced, and the multi-perspective knowledge graph is continuously optimized through user evaluation and feedback, thereby improving the performance of the question-answering system.

6. A question-answering system based on a multi-view knowledge graph, characterized by: It includes a multi-perspective knowledge graph construction module, an interactive receiving module, a node parsing and grading module, a user level calculation module, an extraction and return module, and a user feedback module; The multi-view knowledge graph construction module is used to construct and optimize the multi-view knowledge graph; The interactive receiving module is used to conduct interactive multi-round dialogues with users, receive questions from users, and receive user evaluations and feedback; The node parsing and grading module is used to parse and obtain the corresponding grades of the nodes involved in each round of dialogue; The user level calculation module is used to calculate the user level according to formula (1); The extraction and return module is used to search and obtain answers to questions raised by users in the knowledge graph from the perspective of the corresponding level; The user feedback module is used to obtain user evaluations and feedback, and continuously optimize the multi-view knowledge graph.

Images