CN111767385A

CN111767385A - Intelligent question and answer method and device

Info

Publication number: CN111767385A
Application number: CN202010655836.0A
Authority: CN
Inventors: 袁大为
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-07-09
Filing date: 2020-07-09
Publication date: 2020-10-13

Abstract

The embodiment of the application discloses an intelligent question-answering method and device in the field of artificial intelligent natural language processing; the embodiment of the application is related to technologies such as query and modification of a database, and the embodiment of the application acquires entity type information of a knowledge graph to be constructed; extracting information of the text data based on the entity type information to obtain a plurality of initial entities; synonymy expansion is carried out on each initial entity to obtain a plurality of synonyms of each initial entity, and a plurality of reference entities corresponding to each entity type information are obtained; when a target reference entity matched with the synonym exists, calculating the similarity between a target initial entity corresponding to the synonym and the target reference entity; when the similarity meets a preset condition, performing knowledge fusion based on the target initial entity and the target reference entity to construct a target knowledge graph; outputting answer information of the query information based on the target knowledge graph; the scheme can improve the question-answering accuracy of intelligent question answering.

Description

Intelligent question and answer method and device

Technical Field

The application relates to the field of artificial intelligence, in particular to an intelligent question answering method and device.

Background

The intelligent question-answering technology is a new customer service form, brings convenience to daily life of people to a certain extent, the realization of the intelligent question-answering technology usually depends on mass data stored in advance, in the prior art, the data sorting, data entry and other processes need to be finished manually when data are stored, most of data in a database are question-answer pairs with fixed contents, and in the research and practice processes of the prior art, the inventor of the application finds that the problem that the accuracy rate of the existing intelligent question-answering technology is low due to the fact that more manpower is needed to participate in the existing intelligent question-answering technology, and the matching degree of the question-answer pairs with the fixed contents and the question information received in a practical scene is low.

Disclosure of Invention

The embodiment of the application provides an intelligent question and answer method and device, which can improve the question and answer accuracy of intelligent question and answer.

The embodiment of the application provides an intelligent question answering method, which comprises the following steps:

acquiring entity range information of a knowledge graph to be constructed, wherein the entity range information comprises entity type information;

extracting information of the text data based on the entity type information to obtain initial key information, wherein the initial key information comprises a plurality of initial entities;

synonymy expanding is carried out on each initial entity to obtain a plurality of synonyms of each initial entity, and a plurality of reference entities corresponding to each entity type information are obtained, wherein the initial entities and the synonyms have the same meaning;

when a target reference entity matched with the synonym exists, calculating the similarity between a target initial entity corresponding to the synonym and the target reference entity;

when the similarity meets a preset condition, performing knowledge fusion on the basis of the target initial entity and the target reference entity to construct a target knowledge graph;

and outputting answer information of the inquiry information based on the target knowledge graph.

Correspondingly, the embodiment of the present application provides an intelligent question answering device, including:

the acquisition module is used for acquiring entity range information of the knowledge graph to be constructed, wherein the entity range information comprises entity type information;

the extraction module is used for extracting information of the text data based on the entity type information to obtain initial key information, and the initial key information comprises a plurality of initial entities;

the expansion module is used for performing synonymous expansion on each initial entity to obtain a plurality of synonymous items of each initial entity and a plurality of reference entities corresponding to each entity type information, wherein the initial entities and the synonymous items have the same meaning;

the calculation module is used for calculating the similarity between a target initial entity corresponding to a synonym and a target reference entity when the target reference entity matched with the synonym exists;

the fusion module is used for performing knowledge fusion on the basis of the target initial entity and the target reference entity to construct a target knowledge graph when the similarity meets a preset condition;

and the output module is used for outputting answer information of the inquiry information based on the target knowledge graph.

In some embodiments of the present application, the expansion module includes an expansion sub-module and an acquisition sub-module, wherein,

the expansion submodule is used for carrying out synonymous expansion on each initial entity to obtain a plurality of synonyms of each initial entity, wherein the initial entity and the synonyms have the same meaning;

and the obtaining submodule is used for obtaining a plurality of reference entities corresponding to each entity type information.

In some embodiments of the present application, the obtaining sub-module is specifically configured to:

obtaining a plurality of reference entities corresponding to the entity type information from the existing knowledge graph

At this time, the fusion module is specifically configured to:

performing entity linking based on the target initial entity and the target reference entity to construct an initial knowledge graph;

and combining the initial knowledge graph and the existing knowledge graph to obtain a target knowledge graph.

In some embodiments of the present application, the extraction module comprises a word segmentation sub-module and an extraction sub-module, wherein,

a word segmentation submodule for performing word segmentation processing on the text data to obtain a plurality of word data

And the extraction submodule is used for extracting information of a plurality of word data of the text data based on the extraction mode corresponding to the entity type information to obtain initial key information.

In some embodiments of the present application, the output module includes a parsing submodule, a determination submodule, and an output submodule, wherein,

the analysis submodule is used for analyzing the query information input by the query object to obtain a query keyword;

a determination submodule, configured to determine answer information of the query information based on the target knowledge map and the query keyword;

and the output sub-module is used for outputting answer information of the query information to the query object.

In some embodiments of the present application, the target knowledge-graph includes a plurality of entities and interrelations between the entities, each entity corresponds to a plurality of attributes, and the determining sub-module is specifically configured to:

acquiring a target entity matched with the query keyword from the target knowledge graph;

and determining answer information of the inquiry information based on the mutual relation including the target entities and a plurality of attributes corresponding to the target entities.

In some embodiments of the present application, the intelligent question answering device further comprises:

the determining module is used for determining a relation entity which has a mutual relation with the target entity according to the mutual relation between the entities in the target knowledge graph;

and the display module is used for displaying the query information related to the query information to the query object based on the plurality of attributes of the relationship entity and the target entity.

the current emotion module is used for calculating the current emotion value of the questioning subject based on the negative feedback when the negative feedback of the questioning subject for the answer information is received;

and the switching module is used for switching a manual inquiry channel for the questioning object when the current emotion value is lower than a preset threshold value.

In some embodiments of the present application, the current emotion module includes an acquisition sub-module, an analysis sub-module, and a current emotion sub-module, wherein,

the acquisition sub-module is used for acquiring the historical query record of the query object;

the analysis submodule is used for carrying out emotion analysis on the historical query record to obtain a historical emotion value of the questioning object;

and the current emotion submodule is used for calculating the current emotion value of the questioning object based on the emotion value corresponding to the negative feedback and the historical emotion value.

and the correction module is used for correcting the target knowledge graph based on the negative feedback when the negative feedback of the questioning object for the answer information is received.

Correspondingly, the embodiment of the present application further provides a storage medium, where the storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute any one of the intelligent question answering methods provided in the embodiment of the present application.

Correspondingly, the embodiment of the present application further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and capable of running on the processor, wherein when the processor executes the computer program, any one of the intelligent question answering methods provided by the embodiment of the present application is implemented.

The intelligent question-answering method can automatically carry out a series of processing on text data to obtain the target knowledge graph, compared with the prior art, a large amount of manpower is not needed any more, the intelligent degree of the intelligent question-answering technology is greatly improved, meanwhile, compared with the question-answering pair with fixed content in the prior art, the target knowledge graph can contain richer information quantity, for example, a plurality of attributes of a single main body, a plurality of mutual relations containing the main body and the like, the answer information of the question information can be returned to the question object based on the target knowledge graph, the related question information with the relation to the question information can be returned to the question object based on the knowledge graph, the intelligent degree of the intelligent question-answering technology is greatly improved, and the question accuracy of intelligent question-answering is obviously improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic view of a scenario of an intelligent question answering device provided in an embodiment of the present application;

fig. 2 is a schematic flowchart of an intelligent question answering method provided in an embodiment of the present application;

FIG. 3 is another schematic flow chart diagram illustrating an intelligent question answering method according to an embodiment of the present application;

FIG. 4 is a schematic diagram of an architecture of an intelligent customer service system provided in an embodiment of the present application;

FIG. 5 is an exemplary diagram of a knowledge-graph provided by an embodiment of the present application;

FIG. 6 is an architectural diagram of an atlas construction platform provided in an embodiment of the present application;

FIG. 7 is an interaction diagram of an intelligent question answering method according to an embodiment of the present application;

FIG. 8 is another schematic interaction diagram of the intelligent question answering method according to the embodiment of the present application;

fig. 9 is a schematic structural diagram of an intelligent question answering device provided in an embodiment of the present application;

fig. 10 is another schematic structural diagram of the intelligent question answering device provided in the embodiment of the present application;

fig. 11 is a schematic structural diagram of a computer device provided in an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application, and it is obvious that the embodiments described in the present application are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Artificial Intelligence (AI) is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human Intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Natural Language Processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that enable efficient communication between humans and computers using natural language. Natural language processing is a science integrating linguistics, computer science and mathematics. Therefore, the research in this field will involve natural language, i.e. the language that people use everyday, so it is closely related to the research of linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic question and answer, knowledge mapping, and the like.

The process of constructing the target knowledge graph and the process of outputting the answer information of the query information based on the target knowledge graph relate to artificial intelligence natural language processing and other technologies, the construction of the target knowledge graph and the obtaining of the answer information of the query information can be achieved through the artificial intelligence natural language processing technology, and specific contents are explained in detail through the embodiment of the application.

The embodiment of the application provides an intelligent question answering method and device. Specifically, the embodiment of the present application may be integrated in an intelligent question and answer apparatus, which may be integrated in an intelligent question and answer computer device, where the intelligent question and answer computer device may be an electronic device such as a terminal or a server, and the server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud computing services. The terminal may be, but is not limited to, a smart wearable device, a smart home device, a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and the like. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein.

As shown in fig. 1, fig. 1 is a schematic view of a scenario of an intelligent question answering device according to an embodiment of the present application.

As shown in fig. 1, the server may obtain entity range information of a knowledge graph to be constructed, the entity range information includes entity type information, information extraction is performed on text data based on the entity type information to obtain initial key information, the initial key information includes a plurality of initial entities, synonymy expansion is performed on each initial entity to obtain a plurality of synonyms of each initial entity, and a plurality of reference entities corresponding to each entity type information are obtained, wherein the initial entities and the synonyms thereof have the same meaning, when a target reference entity matching with the synonyms exists, the similarity between the target initial entity corresponding to the synonyms and the target reference entity is calculated, when the similarity satisfies a preset condition, the target knowledge graph is constructed based on the target initial entity and the target reference entity for knowledge fusion, the terminal may send query information to the server, the server may output answer information for the query information based on the target knowledge graph.

It should be noted that the scenario diagram of the intelligent question answering device shown in fig. 1 is only an example, and the intelligent question answering device and the scenario described in the embodiment of the present application are for more clearly illustrating the technical solution of the embodiment of the present application, and do not form a limitation on the technical solution provided in the embodiment of the present application.

The following are detailed below. In this embodiment, description will be made from the perspective of an intelligent question-answering device, which may be integrated on a terminal or a server, as shown in fig. 2, where fig. 2 is a schematic flow chart of an intelligent question-answering method provided in this embodiment of the present application. The intelligent question answering method can comprise the following steps:

101. and acquiring entity range information of the knowledge graph to be constructed, wherein the entity range information comprises entity type information.

A knowledge-graph can be a way to formally describe real-world things and the interrelationships between things.

The entity scope information may include information related to the scope of the things described by the knowledge graph to be constructed, for example, the entity scope of the knowledge graph A may be movies, countries, tourism, entertainment, finance, and the like. The entity scope information defines a scope for the knowledge graph to be constructed, so that the related operation of the knowledge graph to be constructed subsequently can be based on the entity scope information to finally obtain the target knowledge graph which accords with the expectation. Entity scope information may correspond to a data model within a domain, and this data model may contain relevant definition information within the domain, such as entity type information, etc. When the entity scope of the knowledge graph to be constructed is multiple, each entity scope may include the corresponding entity scope information.

The entity type information may be an abstract set of things described by the knowledge graph to be constructed, for example, the entity type information may include director, movie, and movie type.

Acquiring entity range information of the knowledge graph to be constructed, wherein the entity range information can be input manually, such as technicians for leading the construction of the knowledge graph; the entity scope information may also be preset by the system, for example, the system may perform preliminary Semantic understanding (Semantic understating) on the text data of the knowledge graph to be constructed, and automatically determine the entity scope information for the knowledge graph to be constructed according to the result of the Semantic understanding, and the like.

For example, the entity scope of the knowledge graph to be constructed may be "festival," and may receive entity scope information input by a technician and related to the ancient chinese history, where the entity scope information may include entity type information, and the entity type information related to the ancient chinese history may include international festival, traditional festival, celebration mode, and the like.

102. And extracting information of the text data based on the entity type information to obtain initial key information, wherein the initial key information comprises a plurality of initial entities.

The textual material may be raw textual data that includes data used to construct the knowledge-graph, and may be obtained from, for example, encyclopedia, professional literature, official websites, news, etc., in order to ensure the accuracy of the constructed target knowledge-graph.

The initial key information may include content obtained by extracting information from the text material, and the initial key information may be basic knowledge elements, such as initial entities, attributes, and the like.

An entity may be a description of objective content, an entity may be a person, an object, a phenomenon, etc. The initial entity may be a word or sentence or the like extracted from the textual material to describe the objective content.

Based on the entity type information, the process of extracting information from the text data may include entity extraction (NER), relationship extraction, and attribute extraction, where the entity extraction may be to identify entities from the text data, the relationship extraction may be to identify relationships between the entities, the attribute extraction may be to determine a plurality of attributes, the attributes may enrich description of the entities, and the information extraction may be implemented in various ways, for example, the text data may be input into a trained neural network model to obtain initial key information, or for example, the text data may be extracted based on set key information or set extraction rules, and so on.

For example, the initial key information may be obtained by extracting information from text data related to "chinese traditional festival" in encyclopedic through a neural network model, and the initial key information may include a plurality of initial entities (such as spring festival, rice dumpling, etc.).

In some embodiments, the step of extracting information from the text material based on the entity type information to obtain the initial key information may include:

performing word segmentation processing on the text data to obtain a plurality of word data; and extracting information of a plurality of word data of the text data based on the extraction mode corresponding to the entity type information to obtain initial key information.

The word segmentation can be carried out through a word segmentation tool (such as a word segmentation result), after word data are obtained, the word data can be evaluated and screened according to an extraction mode corresponding to the entity type information, so that initial key information in the word data can be determined.

For example, an extraction mode corresponding to "holiday" may be obtained, the text data may be segmented to obtain a plurality of word data, and then the word data may be screened according to the extraction mode to finally obtain a plurality of initial key information.

103. And carrying out synonymy expansion on each initial entity to obtain a plurality of synonyms of each initial entity, and obtaining a plurality of reference entities corresponding to each entity type information, wherein the initial entity and the synonyms have the same meaning.

The reference entity may be a part of concrete matters contained in the entity type information, the entity type information is an abstract description obtained by generalizing the commonality of the reference entity, the reference entity belongs to the entity reference type corresponding to the reference entity, and the initial entity is a character string extracted from the text data and needs to be further processed to determine whether the initial entity belongs to the specific entity type information. The reference entity may be set manually, or may be obtained from a knowledge-graph of the same entity range, and so on.

The description of a certain event may include multiple descriptions, in some situations, descriptions with different meanings may also represent the same meaning, in order to make full use of information in text data as much as possible to construct a more complete and rich knowledge graph, synonymy expansion may be performed on an initial entity to obtain several synonymy terms of the initial entity, and the synonymy expansion may be based on a dictionary (such as a synonym forest), a neural network model (such as a word vector model), or may be expanded by related personnel (such as developers). Specifically, the selection can be flexibly selected according to actual requirements, which is not described herein in detail.

For example, the initial entity "rice dumpling" may be synonymously expanded based on the synonym forest, and the synonyms "rice dumpling, floating rice dumpling, rice dumpling" may be obtained, and the reference entity "rice dumpling, dragon boat, rice dumpling, moon" for obtaining the entity type information "celebration way" may be obtained.

104. And when the target reference entity matched with the synonym exists, calculating the similarity between the target initial entity corresponding to the synonym and the target reference entity.

The target reference entity may be a reference entity matched with the initial entity or the synonym of the initial entity, and the criterion for determining whether to match may be set and adjusted according to an actual scenario, for example, the matching criterion may be that the synonym of the initial entity or the initial entity is completely the same as the reference entity, at this time, the initial entity is the target initial entity, and the reference entity is the target reference entity.

The similarity may include information describing a degree of similarity of the target initial entity to the target reference entity. The similarity may be expressed by characters, for example, by arabic numerals, and the like. Calculating the similarity of the target initial entity and the target reference entity may be performed based on clustering, aggregation, or neural network models, etc.

For example, if it is detected that the synonyms "sweet dumpling" of the initial entity "sweet dumpling" are the same as the "sweet dumpling" of the reference entity, the similarity M between the target initial entity "sweet dumpling" and the target reference entity "sweet dumpling" can be calculated.

105. And when the similarity meets the preset condition, performing knowledge fusion based on the target initial entity and the target reference entity to construct a target knowledge graph.

The preset conditions can determine whether the similarity degree of the target initial entity and the target reference entity is required, the preset conditions can be flexibly set by combining entity range information of the knowledge graph to be constructed, actual requirements, use scenes of the knowledge graph to be constructed and the like, the preset conditions correspond to the similarity, and the expression mode of the preset conditions needs to be set based on the expression mode of the similarity. For example, the predetermined condition may be a predetermined threshold.

The target reference entity may include a plurality of target initial entities corresponding to the target reference entity, and a target reference entity may include a plurality of similarities with the plurality of target initial entities.

In the process of constructing the target knowledge graph, the number of the entity type information, the initial entity, the synonyms of the initial entity, the reference entity and other elements can be multiple, so that a large number of target initial entities with similarity meeting preset conditions and corresponding target reference entities need to be fused, and the target knowledge graph is finally constructed. The knowledge fusion can simplify and integrate scattered and redundant information to obtain a high-quality and rich target knowledge map.

The knowledge fusion can comprise entity linking, namely establishing the corresponding relation between the target initial entity and the target reference entity, and obtaining the target knowledge graph after completing the entity linking of all the initial entities of the text data.

For example, when the similarity M between the target initial entity "sweet dumpling" and the target reference entity "sweet dumpling" matches the preset threshold a, the target knowledge map may be constructed based on the knowledge fusion between the target initial entity "sweet dumpling" and the target reference entity "sweet dumpling".

In some embodiments, the step of "obtaining a number of reference entities corresponding to each entity type information" may include:

acquiring a plurality of reference entities corresponding to entity type information from an existing knowledge graph;

at this time, the step of performing knowledge fusion based on the target initial entity and the target reference entity to construct the target knowledge graph may include:

performing entity linking based on the target initial entity and the target reference entity to construct an initial knowledge graph; and combining the initial knowledge graph with the existing knowledge graph to obtain the target knowledge graph.

The construction of the target knowledge graph can be the updating of the existing knowledge graph, so that the reference entity corresponding to each entity type information can be obtained from the existing knowledge graph.

The process of knowledge fusion may include: and performing entity link on the target initial entity and the target reference entity with the similarity meeting the preset requirement, namely establishing the corresponding relation between the target initial entity and the target reference entity, obtaining an initial knowledge graph based on the text data after completing the entity link of all the initial entities in the text data, and then combining the initial knowledge graph and the existing knowledge graph to obtain the target knowledge graph, namely completing the update of the existing knowledge graph.

For example, reference entities "sweet dumpling, dragon boat, rice dumpling and moon" of entity type information "celebration way" are obtained from the existing knowledge graph 1, then, the corresponding relation between all target initial entities and target reference entities can be established to obtain an initial knowledge graph 2, and then the existing knowledge graph 1 and the initial knowledge graph 2 are combined to obtain a target knowledge graph.

106. Answer information of the query information is output based on the target knowledge graph.

For example, the query message may be "what to eat overnight", and the answer message "rice dumpling" to the query message may be output based on the target knowledge map.

In some embodiments, the step of outputting answer information of the query information based on the target knowledge graph may include:

analyzing the query information input by the query object to obtain query keywords; determining answer information of the query information based on the target knowledge graph and the query keywords; answer information of the query information is output to the questioning object.

The query keywords can comprise texts generally describing the meaning of the query information, and the query keywords can be obtained by segmenting the query information and screening words obtained by segmenting the words; or semantically understand the query information and select a query keyword matched with the query information based on the understood meaning, and the like. Compared with the inquiry information, the inquiry keyword can more normatively and briefly describe the meaning of the inquiry information expression, so that the answer information required by the questioning object is more efficiently and accurately output by using the target knowledge graph.

For example, the query information "what was eaten at noon" input by the query object in the small white is analyzed to obtain the query keyword B, then the answer information "zongzi" of the query information is obtained based on the target knowledge map and the query keyword, and finally the answer information "zongzi" is displayed to the query object in the small white.

In some embodiments, the target knowledge-graph includes a number of entities, each entity corresponding to a number of attributes,

the step of determining answer information of the query information based on the target knowledge graph and the query keyword may include:

acquiring a target entity matched with the query keyword from the target knowledge graph; and determining answer information of the query information based on the correlation containing the target entities and a plurality of attributes corresponding to the target entities.

The knowledge-graph may include a plurality of basic elements, the basic elements may include a plurality of entities, the mutual relations among the entities, each entity contains a plurality of attributes and attribute values corresponding to the attributes, for example, the relation between the entity "moon cake" and the entity "mid-autumn festival" may be "festival feature food", for example, one attribute and attribute value of the entity "mid-autumn festival" may be "date" and "lunar calendar eighty-five month".

The process of determining answer information to the query information may include: and determining a target entity matched with the query keyword in the target knowledge graph, wherein the matching standard can be flexibly set, for example, the matching standard can be that the query keyword is completely the same as an element in the target knowledge graph, and the entity of the element is the target entity. To ensure a more accurate and efficient match, some close or synonymous expansion of elements may be performed, normalized replacement of query keywords (e.g., what to eat is replaced with food), and so on.

For example, the query keyword of the query information "what was eaten at noon festival" may be "noon festival" and "food", then the target entity "noon festival" matching the query keyword may be acquired from the target knowledge map, and then the answer information "rice dumpling" of the query information may be determined according to the attribute, the correlation, and the like of the target entity "noon festival".

In some embodiments, the intelligent question answering method may further include:

determining a relation entity which has a mutual relation with a target entity according to the mutual relation between the entities in the target knowledge graph; query information relevant to the query information is displayed to the query object based on a number of attributes of the relational entity and the target entity.

The related query information can be query information related to meaning expressed by the query information, and in order to improve the intelligence degree of intelligent answering, after the query information is received, the related query information of the query information can be returned based on the query information.

For example, the relational entities that have a relationship with the target entity "end noon festival" may include "rice dumpling", "dragon boat", "wormwood", "yiyuan", and the like, and the attributes of the target entity "end noon festival" may include "date", "meaning", "source", and the like, that is, based on the information such as these relational entities and attributes of the target entity "end noon festival", the query information related to the query information may be displayed to the query object.

when negative feedback of the questioning subject for answer information is received, calculating the current emotion value of the questioning subject based on the negative feedback; and when the current emotion value is lower than a preset threshold value, switching a manual inquiry channel for the questioning object.

The negative feedback may represent that the questioning object does not agree with answer information output based on the knowledge graph, in this embodiment of the present application, a manual query channel may be automatically switched to the questioning object, specifically, the present application may receive feedback information of the output answer information, where the feedback information is a opinion of the questioning object on the answer information, and the feedback information may include the negative feedback.

The current emotion value can be an estimated value of the current mood of the questioning object, and the current emotion value can refer to the influence of the computer equipment integrated with the intelligent question-answering method on the mood of the questioning object. The negative feedback may include different levels, the negative feedback of different levels may correspond to different emotion values, and when the negative feedback is received, the current emotion value of the questioning subject may be determined based on the emotion value corresponding to the negative feedback.

In addition, if the answer information of the query information cannot be obtained based on the knowledge graph, the query object is not satisfied with the computer equipment which integrates and watches the intelligent query and answer method, the emotion of the query object is worsened, a no-output emotion value is generated based on the emotion, the no-output emotion value is equivalent to the degree of the worsening of the emotion state of the query object caused by the fact that the answer information cannot be output, and the current emotion value of the query object is determined based on the no-output emotion value.

If the current emotion value is lower than the preset threshold value, a manual inquiry channel is switched to the questioning object, so that the questioning object can quickly obtain a desired answer, and good use experience is achieved. Before the manual query channel is switched, the query type can be determined according to the query information of the query object, and the manual query channel corresponding to the query type can be switched for the query object.

For example, when negative feedback of the small query subject on answer information is received, the current emotion value 65 of the query subject is calculated based on the negative feedback, and as the current emotion value 65 is lower than the preset threshold value 75, a manual query channel is switched to the small query subject, so that the small query subject can quickly find the desired answer information.

In some embodiments, the step of "calculating the current emotional value of the questioning subject based on negative feedback" may include:

acquiring a historical query record of a query object; performing emotion analysis on the historical query records to obtain historical emotion values of the questioning objects; and calculating the current emotion value of the questioning object based on the emotion value corresponding to the negative feedback and the historical emotion value.

The historical query record can be query information of a query object before the query information, the historical emotion value of the query object can be obtained by analyzing the historical query record of the query object, the historical query record after a certain time node can be obtained according to actual requirements, emotion analysis is carried out on each historical query record to obtain an emotion value, different weights are matched for each emotion value according to the time node corresponding to the emotion value, and then the historical emotion value of the query object is obtained; in addition, the variation range of the emotion value of each historical query record along with the time node can also be used as one of the reference factors for analyzing the emotional value of the questioning object, the variation range can be equivalent to the emotional stability of the questioning object, and the lower the emotional stability is, the easier the emotion of the questioning object changes.

The current emotion value of the questioning object can be obtained by calculating according to the emotion value and the historical emotion value corresponding to the negative feedback, and different weights can be distributed to the emotion value and the historical emotion value corresponding to the negative feedback according to actual conditions.

and when negative feedback of the questioning object for the answer information is received, correcting the target knowledge graph based on the negative feedback.

For example, when negative feedback of the questioning object for the answer information is received, the target entity corresponding to the answer information in the knowledge graph may be located, and the target entity and the content related to the target entity may be checked, and the part having the error may be modified.

The method can acquire the entity range information of the knowledge graph to be constructed, wherein the entity range information comprises entity type information, and then based on the entity type information, extracting the text data to obtain initial key information including several initial entities, then, synonymy expansion is carried out on each initial entity to obtain a plurality of synonymy items of each initial entity, and a plurality of reference entities corresponding to each entity type information are obtained, wherein, the initial entity has the same meaning with the corresponding terms, when a target reference entity matched with the corresponding terms exists, the similarity between the target initial entity corresponding to the corresponding terms and the target reference entity is calculated, when the similarity meets the preset condition, performing knowledge fusion based on the target initial entity and the target reference entity, to construct a target knowledge graph, and finally, to output answer information of the query information based on the target knowledge graph.

In this embodiment, as shown in fig. 3, fig. 3 is a schematic flow chart of the intelligent question answering method provided in the embodiment of the present application. The intelligent question answering method can comprise the following steps:

201. the server acquires entity range information of the knowledge graph to be constructed, wherein the entity range information comprises entity type information.

202. And the server extracts the information of the text data based on the entity type information to obtain initial key information, wherein the initial key information comprises a plurality of initial entities.

203. The server carries out synonymy expansion on each initial entity to obtain a plurality of synonyms of each initial entity and a plurality of reference entities corresponding to each entity type information, wherein the initial entities and the synonyms have the same meaning.

204. When the target reference entity matched with the synonym exists, the server calculates the similarity of the target initial entity corresponding to the synonym and the target reference entity.

205. And when the similarity meets the preset condition, the server performs knowledge fusion on the basis of the target initial entity and the target reference entity to construct a target knowledge graph.

206. The terminal analyzes the query information input by the query object to obtain a query keyword.

207. And the server determines answer information of the inquiry information based on the target knowledge graph and the inquiry keywords and returns the answer information of the inquiry information to the terminal.

208. And when the terminal receives negative feedback of the questioning object for the answer information, calculating the current emotion value of the questioning object based on the negative feedback.

209. And if the current emotion value is lower than the preset threshold value, the terminal switches the manual inquiry channel for the questioning object.

In order to better implement the intelligent question-answering method provided by the embodiment of the application, the embodiment of the application also provides an intelligent customer service system based on the integrated intelligent question-answering method, the overall architecture diagram of the intelligent customer service system can be shown in fig. 4, as shown in fig. 4, the intelligent customer service system can comprise a knowledge base management platform, the knowledge base management platform can comprise a question-answering knowledge base and a map knowledge base, the question-answering knowledge base can input question-answering data in a question-answering pair mode, the question-answering knowledge base can support input of a single question-answering pair, input of a plurality of question-answering pairs in batch mode and the like, the knowledge base can be automatically constructed, the input question-answering pairs are saved in the question-answering knowledge base, and the question-answering pairs in the diagram are a database for saving contents in the question-answering knowledge base.

The knowledge graph knowledge base can automatically construct uploaded text data into a knowledge graph, the obtained knowledge graph can be stored in a graph database, and the knowledge graph can be directly seen on a knowledge graph management platform (as shown in figure 5), specifically, the knowledge graph base can comprise a graph construction platform which is shown in figure 6, the graph construction platform can firstly upload business data, after the business data is received, the business data is intelligently extracted to obtain initial key information, the initial key information comprises initial entities, the relation among the initial entities, the attributes of the initial entities and the like, the initial key information can be stored in a temporary database, then, knowledge cleaning, attribute mapping and knowledge fusion can be carried out based on the initial key information to obtain an initial knowledge graph base, the processes comprise a part which is manually edited, and after the initial knowledge graph is checked, and obtaining the target knowledge graph, and storing the target knowledge graph in a graph database.

Specifically, the knowledge graph creator can define the domain of the knowledge graph to be constructed (the domain can be expressed by defining types and attributes), upload related original corpora, the original corpora can include a plurality of data, documents and the like, the graph management platform can extract knowledge from the related data and documents to obtain a plurality of basic elements, the basic elements can include entities, attributes and interrelations among the entities, the knowledge extraction is mainly open-oriented link data, the available knowledge units are extracted through an automatic technology, the knowledge units mainly include 3 knowledge elements of the entities, the relationships and the attributes, and on the basis of the knowledge elements, a series of high-quality fact expressions are formed to lay the foundation for constructing an upper model layer, specifically, the knowledge extraction can include entity extraction, relationship extraction and attribute extraction, wherein the entity extraction, namely Named Entity Recognition (NER), which is the automatic recognition of named entities from the original corpus; the relation extraction is to solve the problem of semantic link between entities, and mainly identifies entity relation through a relation model between the entities; the attribute extraction is to the entity, a complete sketch of the entity can be formed through the attribute, and the attribute of the entity can be regarded as a name relationship between the entity and the attribute value, so that the extraction problem of the entity attribute can be converted into a relationship extraction problem.

Then, processing the obtained basic elements, including attribute mapping, knowledge clearness and knowledge fusion, and finally obtaining a constructed target knowledge graph which is stored in a graph database, wherein the attribute mapping is to integrate and map similar and repeated attributes in the initial elements onto defined attributes; the knowledge cleaning can be to process the content of the initial element, such as numerical data normalization, formulated content replacement, punctuation mark correction and the like, and clear rules can be configured manually; the knowledge fusion can be to fuse the fragmentary basic elements after emotion into complete graph data, firstly, entity linkage is carried out, namely, the mapping relation between the clear basic elements and entities in the existing database is established, after the entity linkage is completed to obtain an initial knowledge graph, the initial knowledge graph and the existing knowledge graph are merged to obtain the target knowledge graph.

The intelligent customer service system also comprises an intelligent question-answering system, wherein the intelligent question-answering system is based on a question-answer number library and a graph database, and can support question-answer pair retrieval, graph retrieval and multi-turn dialogue mode to return question answers and dialogue guidance based on a dialogue tree (namely, a knowledge graph is utilized, questions asked by a user are intelligently reasoned, and a recommended question method is given). As shown in fig. 7, when the user asks "entrance ticket of sight spot J", the intelligent question-answering system can display answer information based on the knowledge map, and intelligently infer the question that the user wants to ask by using the knowledge map, and give the recommended questions "business time of sight spot J", "phone of sight spot J", and "web address of sight spot J".

The intelligent customer service system also comprises a natural language processing technology (NLP), a Speech Recognition technology (ASR) and a Speech synthesis technology (TTS, Text to Speech), wherein the NLP can accurately and effectively understand the intention of the user by relying on a plurality of natural language processing technologies such as millions of Chinese corpus accumulation, integrated corpus generalization, template matching, model matching and the like, and the user can be intelligently replied by utilizing an intelligent question-and-answer system; ASR supports user speech input (normal chinese-english recognition, humming recognition, boy note recognition, dialect recognition); TTS can be intelligently broadcasted aiming at the reply content, and gives better question answering experience to users.

The intelligent customer service system also comprises an emotion management system based on NLP, which can accurately identify emotion changes in the question and answer process of the user, timely identify the satisfaction degree of the user according to the current emotion of the user and the question and answer accuracy, automatically switch manual customer service under the conditions of poor question and answer efficiency and reduced user satisfaction degree, and provide 1 to 1 better experience for the user. Specifically, the contents in the database of the intelligent customer service system are classified during entry, and when the query information of the user is received, the answer context of the user is recorded, and the type of the query information of the user is judged and recorded. The determination of the question-answering accuracy can include two aspects, namely, the question which is enough to answer the user; secondly, as shown in fig. 8, when the receiving end receives useless feedback from the user, the emotion management system performs algorithm matching by using a machine learning emotion analysis model and combining question and answer scene information and answer accurate information to obtain current emotion data of the user, if the current emotion data is lower than a system matching threshold value, the intelligent robot automatically switches to a corresponding manual customer service channel according to the recorded user question type, reports the user conversation record to a corresponding customer service person, and the customer service person provides 1 to 1 optimal customer service experience. The intelligent customer service system can determine the inquiry type of the user according to the inquiry record of the user and switch the artificial customer service channel of the corresponding type for the user.

The intelligent customer service system in the embodiment can automatically perform a series of processing on the text data to obtain the target knowledge graph, compared with the prior art, a large amount of manpower is not needed any more, the intelligent degree of the intelligent question-answering technology is greatly improved, meanwhile, compared with the answer pair with fixed content in the prior art, the target knowledge graph can contain richer information quantity, such as a plurality of attributes of a single main body, a plurality of mutual relations containing the main body and the like, so that the method can return answer information of the question information to the question object based on the target knowledge graph, can return related question information with relation to the question information to the question object based on the knowledge graph, and greatly improves the intelligent degree of the intelligent question-answering technology.

In order to better implement the intelligent question-answering method provided by the embodiment of the application, the embodiment of the application also provides a device based on the intelligent question-answering method. The meaning of the noun is the same as that in the above intelligent question answering method, and specific implementation details can refer to the description in the method embodiment.

Fig. 9 is a schematic structural diagram of an intelligent question answering device according to an embodiment of the present application, as shown in fig. 9, where the intelligent question answering device may include an obtaining module 301, an extracting module 302, an expanding module 303, a calculating module 304, a fusing module 305, and an output module 306, where,

an obtaining module 301, configured to obtain entity scope information of a knowledge graph to be constructed, where the entity scope information includes entity type information;

an extracting module 302, configured to perform information extraction on the text data based on the entity type information to obtain initial key information, where the initial key information includes a plurality of initial entities;

an expansion module 303, configured to perform synonymous expansion on each initial entity to obtain a plurality of synonymous items of each initial entity, and obtain a plurality of reference entities corresponding to each entity type information, where the initial entity and the synonymous item have the same meaning;

a calculating module 304, configured to calculate, when there is a target reference entity matching the synonym, a similarity between a target initial entity corresponding to the synonym and the target reference entity;

a fusion module 305, configured to perform knowledge fusion based on the target initial entity and the target reference entity to construct a target knowledge graph when the similarity satisfies a preset condition;

and an output module 306, configured to output answer information of the query information based on the target knowledge graph.

obtaining a plurality of reference entities corresponding to entity type information from an existing knowledge graph

At this time, the fusion module is specifically configured to:

and combining the initial knowledge graph with the existing knowledge graph to obtain the target knowledge graph.

the determining submodule is used for determining answer information of the query information based on the target knowledge graph and the query keywords;

and the output sub-module is used for outputting answer information of the inquiry information to the questioning object.

and determining answer information of the query information based on the correlation containing the target entities and a plurality of attributes corresponding to the target entities.

the determining module is used for determining a relation entity which has a mutual relation with a target entity according to the mutual relation between the entities in the target knowledge graph;

and the display module is used for displaying the related inquiry information of the inquiry information to the inquiry object based on the attributes of the relationship entity and the target entity.

In some embodiments of the present application, referring to fig. 10, the intelligent question answering device further comprises:

a current emotion module 307, configured to, when negative feedback of the questioning subject for the answer information is received, calculate a current emotion value of the questioning subject based on the negative feedback;

and the switching module 308 is configured to switch a manual query channel to the query object when the current emotion value is lower than the preset threshold.

the acquisition submodule is used for acquiring the historical query record of the query object;

the analysis submodule is used for carrying out emotion analysis on the historical inquiry records to obtain a historical emotion value of the questioning object;

In this application, the obtaining module 301 may first obtain entity range information of a knowledge graph to be constructed, the entity range information including entity type information, then the extracting module 302 performs information extraction on text data based on the entity type information to obtain initial key information, the initial key information including a plurality of initial entities, then the expanding module 303 performs synonymy expansion on each initial entity to obtain a plurality of synonyms of each initial entity, and obtains a plurality of reference entities corresponding to each entity type information, wherein the initial entities and the synonyms thereof have the same meaning, when there is a target reference entity matching with the synonyms, the calculating module 304 further calculates similarity between the target initial entity corresponding to the synonyms and the target reference entity, when the similarity satisfies a preset condition, the fusing module 305 performs knowledge fusion based on the target initial entity and the target reference entity, to construct a target knowledge graph, and finally, the output module 306 outputs answer information of the query information based on the target knowledge graph.

In addition, an embodiment of the present application further provides a computer device, where the computer device may be a terminal or a server, as shown in fig. 11, which shows a schematic structural diagram of the computer device according to the embodiment of the present application, and specifically:

the computer device may include components such as a processor 601 of one or more processing cores, memory 602 of one or more computer-readable storage media, a power supply 603, and an input unit 604. Those skilled in the art will appreciate that the computer device architecture illustrated in FIG. 11 is not intended to be limiting of computer devices and may include more or less components than those illustrated, or combinations of certain components, or different arrangements of components. Wherein:

the processor 601 is a control center of the computer device, connects various parts of the whole computer device by using various interfaces and lines, and performs various functions of the computer device and processes data by running or executing software programs and/or modules stored in the memory 602 and calling data stored in the memory 602, thereby monitoring the computer device as a whole. Optionally, processor 601 may include one or more processing cores; preferably, the processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles operating system, user pages, application programs, and the like, and the modem processor mainly handles wireless communication. It will be appreciated that the modem processor described above may not be integrated into the processor 601.

The memory 602 may be used to store software programs and modules, and the processor 601 executes various functional applications and data processing by operating the software programs and modules stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to use of the computer device, and the like. Further, the memory 602 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 602 may also include a memory controller to provide the processor 601 with access to the memory 602.

The computer device further comprises a power supply 603 for supplying power to the various components, and preferably, the power supply 603 is logically connected to the processor 601 through a power management system, so that functions of managing charging, discharging, and power consumption are realized through the power management system. The power supply 603 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

The computer device may also include an input unit 604, the input unit 604 being operable to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the computer device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processor 601 in the computer device loads the executable file corresponding to the process of one or more application programs into the memory 602 according to the following instructions, and the processor 601 runs the application programs stored in the memory 602, thereby implementing various functions as follows:

acquiring entity range information of a knowledge graph to be constructed, wherein the entity range information comprises entity type information; extracting information of the text data based on the entity type information to obtain initial key information, wherein the initial key information comprises a plurality of initial entities; synonymy expanding is carried out on each initial entity to obtain a plurality of synonyms of each initial entity, and a plurality of reference entities corresponding to each entity type information are obtained, wherein the initial entities and the synonyms have the same meaning; when a target reference entity matched with the synonym exists, calculating the similarity between a target initial entity corresponding to the synonym and the target reference entity; when the similarity meets a preset condition, performing knowledge fusion based on the target initial entity and the target reference entity to construct a target knowledge graph; answer information of the query information is output based on the target knowledge graph.

The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.

It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by a computer program, which may be stored in a computer-readable storage medium and loaded and executed by a processor, or by related hardware controlled by the computer program.

To this end, embodiments of the present application further provide a storage medium, in which a computer program is stored, where the computer program can be loaded by a processor to execute the steps in any one of the intelligent question answering methods provided in the embodiments of the present application. For example, the computer program may perform the steps of:

Wherein the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.

Since the computer program stored in the storage medium can execute the steps in any of the intelligent question-answering methods provided in the embodiments of the present application, the beneficial effects that can be achieved by any of the intelligent question-answering methods provided in the embodiments of the present application can be achieved, which are detailed in the foregoing embodiments and will not be described again here.

The above detailed description is given to an intelligent question answering method and device provided by the embodiments of the present application, and a specific example is applied in the description to explain the principle and the implementation of the present application, and the description of the above embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. An intelligent question answering method is characterized by comprising the following steps:

2. The method of claim 1, wherein the obtaining a plurality of reference entities corresponding to each entity type information comprises:

acquiring a plurality of reference entities corresponding to the entity type information from an existing knowledge graph;

performing knowledge fusion on the target initial entity and the target reference entity to construct a target knowledge graph, including:

3. The method of claim 1, wherein the extracting text material information based on the entity type information to obtain initial key information comprises:

performing word segmentation processing on the text data to obtain a plurality of word data;

and extracting information of a plurality of word data of the text data based on the extraction mode corresponding to the entity type information to obtain initial key information.

4. The method of claim 1, wherein outputting answer information for query information based on the target knowledge-graph comprises:

analyzing the query information input by the query object to obtain query keywords;

determining answer information of the query information based on the target knowledge graph and the query keyword;

and outputting answer information of the inquiry information to the questioning object.

5. The method of claim 4, wherein the target knowledge-graph comprises a plurality of entities and interrelationships between the entities, each entity corresponding to a plurality of attributes,

the determining answer information of the query information based on the target knowledge graph and the query keyword comprises:

6. The method of claim 5, further comprising:

determining a relation entity which has a mutual relation with the target entity according to the mutual relation between the entities in the target knowledge graph;

and displaying relevant query information of the query information to the query object based on a plurality of attributes of the relationship entity and the target entity.

7. The method of claim 4, wherein after outputting answer information of the query information based on the target knowledge graph, further comprising:

when negative feedback of the questioning subject for the answer information is received, calculating the current emotion value of the questioning subject based on the negative feedback;

and when the current emotion value is lower than a preset threshold value, switching a manual inquiry channel for the questioning object.

8. The method of claim 7, wherein said calculating a current emotional value of the questioning subject based on the negative feedback comprises:

acquiring a historical query record of the query object;

performing emotion analysis on the historical query record to obtain a historical emotion value of the questioning object;

and calculating the current emotion value of the questioning object based on the emotion value corresponding to the negative feedback and the historical emotion value.

9. The method of claim 7, further comprising:

10. An intelligent question answering device, comprising: