A scalable intelligent question answering method and system

Abstract

The present invention discloses an extensible intelligent question-answering method and system, proposes a functional entity, and expands and associates the functional entity into a knowledge graph system, expands the functional entity on the knowledge graph, and associates user questions with the functional entity through intent recognition, word slot extraction, and entity linking. The method is applied in an intelligent question-answering system to meet the scenario of personalized question-answering of users. When users conduct personalized intention question-answering, they can find specific functional entities through intent recognition, and use the association relationship between the functional entity and other entities in the graph to obtain related data for automatic binding, thereby realizing personalized question-answering, breaking through the traditional knowledge graph that can only conduct graph relationship queries limited to single entity, multiple entities, entity relationships, etc., and giving answers that meet the user's specific intention questions, thereby providing flexible support for personalized question-answering of users.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and more specifically, to a scalable intelligent question-answering method and system. Background Art

[0002] Intelligent question-answering aims to automatically provide answers to natural language questions raised by users. The main technologies currently include intelligent question-answering based on knowledge graphs and intelligent question-answering based on question-answer pairs. The former mainly converts the user's question into entity and predicate relationships, and then obtains the answer corresponding to the question through the relationship in the knowledge graph; the latter mainly matches the user's question with the question in the question-answer pair for text or semantic familiarity, and then selects the best matching answer from the question-answer pair.

[0003] These intelligent question-and-answer technologies have obvious shortcomings in specific business usage scenarios. For example, when a user asks "Is 40 degrees oil temperature of main transformer No. 1 normal?", the question-and-answer based on knowledge graph can extract the "Main Transformer No. 1" entity and its "oil temperature" attribute, but the knowledge graph only records the relationship or attribute information of the entity and cannot answer whether 40 degrees is normal. Question-and-answer based on question-and-answer pairs cannot enumerate all user questions. For example, users may ask "Is 42 degrees oil temperature of main transformer No. 1 normal?", "Is 45 degrees oil temperature of main transformer No. 2 normal?", etc.

[0004] Therefore, a scalable intelligent question answering method is needed. Summary of the invention

[0005] The present invention proposes an extended intelligent question-answering method and system to solve the problem of how to achieve personalized question-answering.

[0006] In order to solve the above problem, according to one aspect of the present invention, a scalable intelligent question answering method is provided, the method comprising:

[0007] Build multiple functional entities based on the user's personalized question-and-answer needs, and set parameter information and training data templates for each functional entity;

[0008] Add each functional entity to the knowledge graph, and establish an association relationship between each functional entity and a related real entity in the knowledge graph according to the parameter information of each functional entity;

[0009] Determine training data according to the training data template and the real entity;

[0010] Using a long short-term memory (LSTM) neural network to train the training data to determine a functional entity intent recognition and word slot extraction model;

[0011] The functional entity intention recognition and word slot extraction module is used to determine the functional entity information and word slot information corresponding to the acquired personalized question, and the answer to the personalized question is determined according to the functional entity information, word slot information and parameter information of the functional entity.

[0012] Preferably, the parameter information of the functional entity includes: the name of each functional entity, the corresponding REST service address URL and the corresponding REST service parameters.

[0013] Preferably, the REST service parameters include: entity relationships or attributes derived from the knowledge graph and dynamic parameters in the personalized questions.

[0014] Preferably, the step of determining the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity comprises:

[0015] The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

[0016] According to another aspect of the present invention, a scalable intelligent question-answering system is provided, the system comprising:

[0017] A functional entity construction unit, used to construct multiple functional entities based on the user's personalized question-and-answer requirements, and set parameter information and training data templates for each functional entity;

[0018] An association relationship establishing unit, used to add each functional entity to the knowledge graph, and establish an association relationship between each functional entity and a related real entity in the knowledge graph according to parameter information of each functional entity;

[0019] A training data determination unit, configured to determine training data according to the training data template and the real entity;

[0020] A model determination unit, used for training the training data using a long short-term memory (LSTM) neural network to determine a functional entity intent recognition and word slot extraction model;

[0021] The answer determination unit is used to determine the functional entity information and word slot information corresponding to the acquired personalized question by using the functional entity intention recognition and word slot extraction module, and determine the answer to the personalized question according to the functional entity information, word slot information and parameter information of the functional entity.

[0022] Preferably, the parameter information of the functional entity includes: the name of each functional entity, the corresponding REST service address URL and the corresponding REST service parameters.

[0023] Preferably, the REST service parameters include: entity relationships or attributes derived from the knowledge graph and dynamic parameters in the personalized questions.

[0024] Preferably, the answer determination unit determines the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity, including:

[0025] The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

[0026] The present invention provides an extensible intelligent question-answering method and system, proposes a functional entity, and expands and associates the functional entity into a knowledge graph system, expands the functional entity on the knowledge graph, and associates user questions with the functional entity through intent recognition, word slot extraction, and entity linking. It is applied in an intelligent question-answering system to meet the scenario of user personalized question-answering. When a user performs personalized intent question-answering, the specific functional entity can be found through intent recognition, and the association relationship between the functional entity and other entities in the graph can be used to obtain related data for automatic binding, thereby realizing personalized question-answering, breaking through the traditional knowledge graph that can only perform graph relationship queries limited to single entity, multiple entities, entity relationships, etc., and giving answers that meet the user's specific intent questions, providing flexible support for user personalized question-answering. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0028] Figure 1 is a flow chart of an expandable intelligent question-answering method 100 according to an embodiment of the present invention;

[0029] Figure 2 A schematic diagram of setting parameter information of a functional entity according to an embodiment of the present invention;

[0030] Figure 3 A schematic diagram of a knowledge graph according to an embodiment of the present invention;

[0031] Figure 4 is a schematic diagram of a training data module according to an embodiment of the present invention;

[0032] Figure 5 A schematic diagram of determining functional entity information and word slot information according to an embodiment of the present invention;

[0033] Figure 6Schematic diagram of the structure of an expandable intelligent question-answering system 600 according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] Now, exemplary embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to disclose the present invention in detail and completely and to fully convey the scope of the present invention to those skilled in the art. The terms used in the exemplary embodiments shown in the accompanying drawings are not intended to limit the present invention. In the accompanying drawings, the same units / elements are marked with the same reference numerals.

[0035] Unless otherwise specified, the terms (including technical terms) used herein have the commonly understood meanings to those skilled in the art. In addition, it is understood that the terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0036] Knowledge graph technology describes concepts, entities and their relationships in the objective world in a structured form, expresses Internet information in a form closer to the human cognitive world, and provides a better ability to organize, manage and understand the massive amount of Internet information. For example, in the field of information technology in the power industry, although the vocabulary of information systems, information applications and APPs is different, the essential meaning they express is the same. If this essential meaning is expressed in a computer model, it is concepts, entities and relationships. This model can be recognized and calculated by computers. Through the knowledge graph KG technology, the abstract cognition existing in the human brain can be converted into a machine-recognizable and computable mathematical model. The knowledge representation route can be divided into two types: semantic web and attribute graph. Among them, the attribute graph model storage components include neo4j, TigerGraph graph database, etc. The semantic web storage components are relatively scattered, and most of them are developed based on open source + NoSQL databases, including Jena TDB + MangoDb, gStoreRDF database, etc.

[0037] LSTM, Long Short Term Memory, is a special RNN network model that can learn long-term dependency information.

[0038] Figure 1 FIG. 1 is a flow chart of an expandable intelligent question-answering method 100 according to an embodiment of the present invention. Figure 1As shown, the scalable intelligent question-answering method and method provided by the embodiment of the present invention proposes a functional entity, and expands and associates the functional entity to the knowledge graph system, expands the functional entity on the knowledge graph, and associates the user's question with the functional entity through intent recognition, word slot extraction, and entity linking, and is applied in the intelligent question-answering system to meet the scenario of user personalized question-answering; when the user conducts personalized intention question-answering, the specific functional entity can be found through intent recognition, and the association relationship between the functional entity and other entities in the graph is used to obtain related data for automatic binding, that is, personalized question-answering can be realized, breaking through the traditional knowledge graph that can only perform single entity, multiple entities, entity relationships, etc., which are limited to graph relationship queries within the knowledge graph, and answers that meet the user's specific intention questions are given, providing flexible support for user personalized question-answering. The scalable intelligent question-answering method 100 provided by the embodiment of the present invention starts from step 101, and in step 101, multiple functional entities are constructed based on the user's personalized question-answering needs, and parameter information and training data templates for each functional entity are set.

[0039] Preferably, the parameter information of the functional entity includes: the name of each functional entity, the corresponding REST service address URL and the corresponding REST service parameters.

[0040] Preferably, the REST service parameters include: entity relationships or attributes derived from the knowledge graph and dynamic parameters in the personalized questions.

[0041] In an embodiment of the present invention, a functional entity is constructed according to the personalized question and answer requirements of the user, and the name of the functional entity (for intent recognition), rest service address URL, rest service parameters and training data template are set. Among them, the rest service parameters include two types, one is the entity relationship or attribute derived from the graph, and the other is the dynamic parameter included by the user in the question. For the training data template, for example: the user's question is "Is the oil temperature of 42 degrees in the No. 1 main transformer normal?", then the labeled training data template structure is "{No. 1 main transformer, transformer entity} {42, numerical value} degrees {oil temperature, transformer attribute, temperature} is normal)". Set the parameter information of the functional entity such as Figure 2 shown.

[0042] In step 102, each functional entity is added to the knowledge graph, and an association relationship between each functional entity and a related real entity in the knowledge graph is established based on parameter information of each functional entity.

[0043] In the embodiment of the present invention, the functional entity is added to the knowledge graph and bound to the relevant real entity in the knowledge graph according to the parameter information of step 101. For example, the transformer oil temperature judgment functional entity is associated with the transformer entity, and the bound knowledge graph is as follows: Figure 3 shown.

[0044] In step 103, training data is determined according to the training data template and the real entity.

[0045] In the implementation of the present invention, after the functional entity is bound to the knowledge graph, a large amount of training data is generated according to the associated entity based on the configured training data template. Figure 4 As shown, according to the template "{transformer entity}{42, value,(^[0-9]{1,3})$} degree {oil temperature, transformer attribute, temperature} is normal" can be combined with the transformer entity bound in the knowledge graph and the set numerical regular expression relationship to generate massive sample data such as: "{main transformer No. 1, transformer entity}{42, value} degree {oil temperature, transformer attribute, temperature} is normal)", "{main transformer No. 2, transformer entity}{55, value} degree {oil temperature, transformer attribute, temperature} is normal)", "{main transformer No. 3, transformer entity}{180, value} degree {oil temperature, transformer attribute, temperature} is normal)" and so on.

[0046] In step 104, a long short-term memory (LSTM) neural network is used to train the training data to determine a functional entity intent recognition and word slot extraction model.

[0047] In the implementation of the present invention, based on massive sample data, LSTM neural network training is used to generate functional entity intention recognition and word slot extraction models. First, it supports intention recognition of user personalized questions (q_template part), which is used to convert user input into a certain functional entity. Second, it extracts the word slot of user personalized questions (parsed part), extracts key information from user input, and performs entity linking. For example, if the user input may be transformer No. 3, then entity linking needs to be performed as main transformer No. 3.

[0048] In step 105, the functional entity intention recognition and word slot extraction module is used to determine the functional entity information and word slot information corresponding to the acquired personalized question, and the answer to the personalized question is determined based on the functional entity information, word slot information and parameter information of the functional entity.

[0049] Preferably, the step of determining the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity comprises:

[0050] The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

[0051] In the embodiment of the present invention, when the user inputs a personalized question, the corresponding functional entity information (q_template) and word slot information (parsed) are found through the above model. The process of determining the functional entity information and word slot information is as follows: Figure 5 After obtaining the functional entity information and word slot information, the rest service parameters set by the functional entity in step 101 are bound, and then the service address URL corresponding to the functional entity setting is called to return the answer required by the user.

[0052] Figure 6 FIG. 6 is a schematic diagram of the structure of an expandable intelligent question-answering system 600 according to an embodiment of the present invention. Figure 6 As shown, the scalable intelligent question-answering system 600 provided by the embodiment of the present invention includes: a functional entity construction unit 601, an association relationship establishment unit 602, a training data determination unit 603, a model determination unit 604 and an answer determination unit 605.

[0053] Preferably, the functional entity construction unit 601 is used to construct multiple functional entities based on the user's personalized question and answer requirements, and set parameter information and training data templates for each functional entity.

[0054] Preferably, the parameter information of the functional entity includes: the name of each functional entity, the corresponding REST service address URL and the corresponding REST service parameters.

[0055] Preferably, the REST service parameters include: entity relationships or attributes derived from the knowledge graph and dynamic parameters in the personalized questions.

[0056] Preferably, the association relationship establishing unit 602 is used to add each functional entity to the knowledge graph, and establish an association relationship between each functional entity and a related real entity in the knowledge graph based on parameter information of each functional entity.

[0057] Preferably, the training data determining unit 603 is used to determine the training data according to the training data template and the real entity.

[0058] Preferably, the model determination unit 604 is used to train the training data using a long short-term memory (LSTM) neural network to determine a functional entity intent recognition and word slot extraction model.

[0059] Preferably, the answer determination unit 605 is used to use the functional entity intention recognition and word slot extraction module to determine the functional entity information and word slot information corresponding to the acquired personalized question, and determine the answer to the personalized question based on the functional entity information, word slot information and parameter information of the functional entity.

[0060] Preferably, the answer determination unit 605 determines the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity, including:

[0061] The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

[0062] The scalable intelligent question answering system 600 of an embodiment of the present invention corresponds to the scalable intelligent question answering method 100 of another embodiment of the present invention, which will not be described in detail here.

[0063] The invention has been described above with reference to a few embodiments. However, it is readily apparent to a person skilled in the art that other embodiments than the ones disclosed above are equally within the scope of the invention, as defined by the appended patent claims.

[0064] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a / said / the [means, components, etc.]" are to be openly interpreted as at least one instance of the means, components, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily have to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0065] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0066] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A scalable intelligent question-answering method, characterized in that: The method comprises: Build multiple functional entities based on the user's personalized question-and-answer needs, and set parameter information and training data templates for each functional entity; Add each functional entity to the knowledge graph, and establish an association relationship between each functional entity and a related real entity in the knowledge graph according to the parameter information of each functional entity; Determine training data according to the training data template and the real entity; The training data is trained using a long short-term memory (LSTM) neural network to determine a functional entity intent recognition and word slot extraction model; Determine the functional entity information and word slot information corresponding to the acquired personalized question by using the functional entity intention recognition and word slot extraction module, and determine the answer to the personalized question according to the functional entity information, word slot information and parameter information of the functional entity; The parameter information of the functional entity includes: the name of each functional entity, the corresponding rest service address URL and the corresponding rest service parameters; Wherein, determining the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity includes: The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

2. The method according to claim 1, characterized in that The rest service parameters include: entity relationships or attributes from the knowledge graph and dynamic parameters in the personalized questions.

3. A scalable intelligent question-answering system, characterized in that: The system comprises: A functional entity construction unit, used to construct multiple functional entities based on the user's personalized question-and-answer requirements, and set parameter information and training data templates for each functional entity; An association relationship establishing unit, used to add each functional entity to the knowledge graph, and establish an association relationship between each functional entity and a related real entity in the knowledge graph according to parameter information of each functional entity; A training data determination unit, configured to determine training data according to the training data template and the real entity; A model determination unit, used for training the training data using a long short-term memory (LSTM) neural network to determine a functional entity intent recognition and word slot extraction model; an answer determination unit, configured to determine the functional entity information and word slot information corresponding to the acquired personalized question by using the functional entity intention recognition and word slot extraction module, and determine the answer to the personalized question according to the functional entity information, word slot information and parameter information of the functional entity; The parameter information of the functional entity includes: the name of each functional entity, the corresponding rest service address URL and the corresponding rest service parameters; The answer determination unit determines the answer to the personalized question according to the functional entity information, the word slot information and the parameter information of the functional entity, including: The rest service parameters of the functional entity are bound according to the word slot information, and the rest service address URL corresponding to the functional entity is called according to the functional entity information to obtain the answer to the personalized question.

4. The system according to claim 3, characterized in that The rest service parameters include: entity relationships or attributes from the knowledge graph and dynamic parameters in the personalized questions.

Images