CN112749263A - Multi-round answer generation system based on single question - Google Patents
Multi-round answer generation system based on single question Download PDFInfo
- Publication number
- CN112749263A CN112749263A CN202011262651.XA CN202011262651A CN112749263A CN 112749263 A CN112749263 A CN 112749263A CN 202011262651 A CN202011262651 A CN 202011262651A CN 112749263 A CN112749263 A CN 112749263A
- Authority
- CN
- China
- Prior art keywords
- module
- natural language
- knowledge
- intelligent
- knowledge base
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3334—Selection or weighting of terms from queries, including natural language queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
- G06F40/35—Discourse or dialogue representation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention relates to the field of man-machine interaction, in particular to a multi-round answer generation system based on a single question, which is based on corresponding map knowledge data and comprises a natural language processing module, a dialogue management module, an intelligent semantic understanding module, an answer intelligent acquisition module and a knowledge base module; by adopting the method, the purpose of conversation of the user is determined through intention identification, key points of user consultation are accurately obtained through entity identification, multiple times of inquiry, confirmation and the like, and the decision and propulsion of each round of conversation are completed through a unique conversation management mechanism, so that multiple rounds of answers are generated based on a certain problem, and the effective and friendly operation of man-machine conversation is ensured; moreover, the success probability of man-machine interaction is further improved, and the problems of difficulty in obtaining user service, traditional service mode and the like are effectively solved.
Description
Technical Field
The invention relates to the field of human-computer interaction, in particular to a multi-round answer generation system based on a single question.
Background
The intelligent question-answering system is an intelligent robot system capable of answering questions in any natural language form, and for the questions to be asked, semantic information analysis is carried out on the questions through a natural language understanding analysis technology and a knowledge graph technology, so that correct answers are found in the contents of a large-scale knowledge base, and a list formed by a plurality of webpages is not returned like a keyword search engine. An important function in intelligent question answering is multiple rounds of interaction based on a context dialog scenario, which is also one of the difficulties.
In practical applications, the intelligent question-answering system is not a simple question-answer but may be complex flow-type knowledge. The conversation management controls the process of man-machine conversation, and determines the current reaction to the user according to the conversation content.
Most of the existing technologies for generating multiple rounds of answers based on a certain question are multiple pre-defined multi-round question-answer action sets; when the system is in operation, according to the questions consulted by the current user, the most optimal system action is selected from a plurality of pre-defined question-answer action sets through a series of strategies to be output. However, the solution is not ideal.
For the intelligent question-answering system with the tree-shaped hierarchical dependency relationship of each system action, the rules for generating multiple rounds of answers based on a certain question are manually defined, the problem customization is complex, and the conflict of multiple rules is easy to occur; the statistical dialogue system based on reinforcement learning can automatically learn the tree-shaped dependency relationship on the premise of sufficient training corpora, but the corpora are difficult to obtain, and the learned content is poor in understandability and difficult to control.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a multi-round answer generation system based on a single question, which is mainly used for carrying out data preparation and model training on different types of questions to obtain intelligent knowledge graph data, providing a multi-round dialogue management mechanism and a decision mechanism and completing intelligent question answering based on a certain question to generate multi-round answers.
In order to achieve the purpose, the invention provides the following technical scheme:
a multi-round answer generation system based on a single question is characterized by comprising a natural language processing module, a dialogue management module, an intelligent semantic understanding module, an answer intelligent acquisition module and a knowledge base module on the basis of corresponding map knowledge data;
the natural language processing module completes semantic analysis on the text and generates a corresponding natural language text aiming at the input of a user; the dialogue management module is used for processing complex problems of context problems; the intelligent semantic understanding module is used for performing semantic understanding on a certain problem; the intelligent answer obtaining module is used for obtaining answers to questions; the knowledge base module is used for constructing and updating a knowledge base;
the natural language processing module is connected with an external signal output device; the dialogue understanding module is communicated with the natural language processing module and the intelligent semantic understanding module; the answer intelligent acquisition module is communicated with the knowledge base module; the knowledge base module is communicated with an external online database.
Further, the natural language processing module is internally divided into a natural language understanding NLU and a natural language generating NLG, wherein the natural language understanding NLU: and completing semantic analysis of the text, and extracting key information such as entities, intents and the like. Natural language generation NLG: generating responsive natural language text for input by a user;
natural Language Processing (NLP) is a field of computer science, artificial intelligence, linguistics that focuses on the interaction between computers and human (natural) language. Thus, natural language processing is relevant to the field of human-computer interaction. Natural language processing faces many challenges, including natural language understanding, and thus, natural language processing involves an area of human-computer interaction. Many challenges in NLP relate to natural language understanding, i.e., computer-derived meaning from human or natural language input, and other concerns with natural language generation.
Furthermore, the dialogue management module realizes the state control and tracking, data management and context management of the dialogue process.
Further, the knowledge base module comprises automatic knowledge acquisition, manual knowledge input, automatic knowledge increment and invalid knowledge deletion, and the completeness and effectiveness of knowledge in the knowledge base are guaranteed;
compared with the prior art, the invention provides a multi-round answer generation system based on a single question, which has the following beneficial effects:
by adopting the method, the intelligent atlas technology for generating multi-turn answers based on a certain question needs each turn of dialogue to clarify the intention and key information of the user, and actively proposes an inquiry to guide the user to select or guide the user to clarify the intention and a question method in the dialogue process of answering the user question; when the user further answers, the intelligent map technology is required to add constraint conditions during semantic understanding so that the intelligent map technology can automatically understand, and therefore the understanding correctness is guaranteed.
The method determines the conversation purpose of the user through intention identification, accurately obtains key points of user consultation through entity identification, multiple times of inquiry, confirmation and the like, and finishes decision and propulsion of each round of conversation through a unique conversation management mechanism, thereby realizing generation of multiple rounds of answers based on a certain problem and ensuring effective and friendly operation of man-machine conversation.
Moreover, the success probability of man-machine interaction is further improved, and the problems of difficulty in obtaining user service, traditional service mode and the like are effectively solved.
Drawings
FIG. 1 is an overall flow diagram of the present invention
Fig. 2 is a schematic diagram of the flow for using feedback in accordance with the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Referring to fig. 1-2, a multi-round answer generation system based on a single question based on corresponding map knowledge data comprises a natural language processing module, a dialogue management module, an intelligent semantic understanding module, an answer intelligent acquisition module and a knowledge base module;
the natural language processing module completes semantic analysis on the text and generates a corresponding natural language text aiming at the input of a user; the dialogue management module is used for processing complex problems of context problems; the intelligent semantic understanding module is used for performing semantic understanding on a certain problem; the intelligent answer obtaining module is used for obtaining answers to questions; the knowledge base module is used for constructing and updating a knowledge base;
the natural language processing module is connected with an external signal output device; the dialogue understanding module is communicated with the natural language processing module and the intelligent semantic understanding module; the answer intelligent acquisition module is communicated with the knowledge base module; the knowledge base module is communicated with an external online database.
In the invention, data preparation is carried out aiming at different problems, input and analysis are carried out through a natural language processing module, intelligent knowledge map data are obtained through a training model, and each turn of conversation is promoted through a multi-turn conversation management mechanism and a decision mechanism in a conversation management module so as to finish the intelligent question-answering method; the meaning indicated by the training model is that an intention classification model and an element extraction model are obtained through training and are used for analyzing the intention and key elements of the input content of the user; the system comprises a multi-turn dialogue management mechanism and a decision mechanism, wherein the multi-turn dialogue management mechanism and the decision mechanism are used for effectively assisting a user in completing multi-turn dialogue around a certain problem by continuously deciding the optimal action to be taken next according to the current problem;
obtaining an intention classification model and an element extraction model in a training process, wherein the training comprises judging whether the intention of a user is transferred or not through logistic regression;
the specific method is that the relevance of the same problem context is comprehensively considered, and cosine similarity and distributed similarity of relevant single features are used as feature vectors of logistic regression classification to judge whether the intention is transferred or not;
the single characteristics of the relevancy comprise TF-IDF, chi-square and information entropy; TF-IDF is a commonly used weighting technique for information retrieval and data mining; TF is the word frequency and IDF is the inverse text frequency index; aiming at the single characteristics, a Bi-LSTM method is adopted for training to carry out entity recognition;
the dialogue management mechanism and the decision mechanism in the dialogue management module are used for establishing a conversation mechanism by taking the element of the user utterance as a drive;
based on a certain question and a plurality of rounds of conversations, when the element user of the question is not clear or the provided data is incomplete, the robot guides the user to complete element providing and acquire the complete element of the question and answer. Firstly, classifying key elements into word slots of entity types, and accurately acquiring consultation key points of a user through word slot recognition, multiple queries and confirmation actions; different questions are classified into different categories of intentions, and the purpose of the user's conversation is identified and determined by the intentions.
In the present invention, the natural language processing module is internally divided into a natural language understanding NLU and a natural language generating NLG, wherein the natural language understanding NLU: and completing semantic analysis of the text, and extracting key information such as entities, intents and the like. Natural language generation NLG: generating responsive natural language text for input by a user; furthermore, the dialogue management module realizes the state control and tracking, data management and context management of the dialogue process.
Further, the knowledge base module comprises automatic knowledge acquisition, manual knowledge input, automatic knowledge increment and invalid knowledge deletion, and the completeness and effectiveness of knowledge in the knowledge base are guaranteed;
in the system, the specific acquisition flow of the answer is that a user proposes a consultation question, and a natural language processing module converts the natural language of the user into a recognizable signal or variable; analyzing the intention and extracting elements through an intelligent semantic understanding module, and feeding back the intention and the elements to a user through a dialogue management module, wherein in the process, for the intention and the entry with a larger range, a system prompts to narrow the range and change the range; in the conversation process, the conversation management module tracks the conversation state all the time, classifies key elements into word slots of entity types, accurately acquires the consultation key points of the user through word slot recognition and multiple inquiry and confirmation actions, and compares the results with subsequent information links corresponding to the word slots in a knowledge base; reducing the application range through a multi-turn dialogue management mechanism and a decision mechanism, and finally obtaining the required answer;
referring to fig. 1-2, with the present invention, an intelligent atlas technique for generating multiple rounds of answers based on a certain question requires each round of dialog to clarify the intention and key information of the user, and actively proposes an inquiry to guide the user to make a selection or to guide the user to clarify the intention and a query method in the dialog process of answering the user question; when the user further answers, the intelligent map technology is required to add constraint conditions during semantic understanding so that the intelligent map technology can automatically understand, and therefore the understanding correctness is guaranteed.
The method determines the conversation purpose of the user through intention identification, accurately obtains key points of user consultation through entity identification, multiple times of inquiry, confirmation and the like, and finishes decision and propulsion of each round of conversation through a unique conversation management mechanism, thereby realizing generation of multiple rounds of answers based on a certain problem and ensuring effective and friendly operation of man-machine conversation.
Moreover, the success probability of man-machine interaction is further improved, and the problems of difficulty in obtaining user service, traditional service mode and the like are effectively solved.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications, additions and substitutions for the described embodiments may be made by those skilled in the art without departing from the scope and spirit of the invention as defined by the accompanying claims.
Claims (4)
1. A multi-round answer generation system based on a single question is characterized by comprising a natural language processing module, a dialogue management module, an intelligent semantic understanding module, an answer intelligent acquisition module and a knowledge base module on the basis of corresponding map knowledge data;
the natural language processing module completes semantic analysis on the text and generates a corresponding natural language text aiming at the input of a user; the dialogue management module is used for processing complex problems of context problems; the intelligent semantic understanding module is used for performing semantic understanding on a certain problem; the intelligent answer obtaining module is used for obtaining answers to questions; the knowledge base module is used for constructing and updating a knowledge base;
the natural language processing module is connected with an external signal output device; the dialogue understanding module is communicated with the natural language processing module and the intelligent semantic understanding module; the answer intelligent acquisition module is communicated with the knowledge base module; the knowledge base module is communicated with an external online database.
2. The single question based multi-round answer generation system of claim 1, wherein the natural language processing module is internally divided into a natural language understanding NLU and a natural language generating NLG, wherein the natural language understanding NLU: and completing semantic analysis of the text, and extracting key information such as entities, intents and the like. Natural language generation NLG: natural language text responsive to user input
3. The system of claim 1, wherein the dialogue management module implements state control and tracking, data management, and context management of the dialogue process.
4. The system of claim 1, wherein the knowledge base module comprises automatic knowledge acquisition, manual knowledge entry, automatic knowledge increment and invalid knowledge deletion, and ensures knowledge integrity and validity of the knowledge base.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011262651.XA CN112749263A (en) | 2020-11-12 | 2020-11-12 | Multi-round answer generation system based on single question |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011262651.XA CN112749263A (en) | 2020-11-12 | 2020-11-12 | Multi-round answer generation system based on single question |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112749263A true CN112749263A (en) | 2021-05-04 |
Family
ID=75648883
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011262651.XA Withdrawn CN112749263A (en) | 2020-11-12 | 2020-11-12 | Multi-round answer generation system based on single question |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112749263A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113553405A (en) * | 2021-06-11 | 2021-10-26 | 中国农业银行股份有限公司浙江省分行 | Chinese-character-bert-model-based intelligent robot implementation method and system |
| CN114020894A (en) * | 2021-11-08 | 2022-02-08 | 桂林电子科技大学 | Intelligent evaluation system capable of realizing multi-round interaction |
| CN117348922A (en) * | 2023-11-06 | 2024-01-05 | 中国海洋大学 | Interactive dialogue generation method for application software configuration of Internet of things |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170228372A1 (en) * | 2016-02-08 | 2017-08-10 | Taiger Spain Sl | System and method for querying questions and answers |
| CN109446306A (en) * | 2018-10-16 | 2019-03-08 | 浪潮软件股份有限公司 | Task-driven multi-turn dialogue-based intelligent question and answer method |
-
2020
- 2020-11-12 CN CN202011262651.XA patent/CN112749263A/en not_active Withdrawn
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170228372A1 (en) * | 2016-02-08 | 2017-08-10 | Taiger Spain Sl | System and method for querying questions and answers |
| CN109446306A (en) * | 2018-10-16 | 2019-03-08 | 浪潮软件股份有限公司 | Task-driven multi-turn dialogue-based intelligent question and answer method |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113553405A (en) * | 2021-06-11 | 2021-10-26 | 中国农业银行股份有限公司浙江省分行 | Chinese-character-bert-model-based intelligent robot implementation method and system |
| CN114020894A (en) * | 2021-11-08 | 2022-02-08 | 桂林电子科技大学 | Intelligent evaluation system capable of realizing multi-round interaction |
| CN114020894B (en) * | 2021-11-08 | 2024-03-26 | 桂林电子科技大学 | Intelligent evaluation system capable of realizing multi-wheel interaction |
| CN117348922A (en) * | 2023-11-06 | 2024-01-05 | 中国海洋大学 | Interactive dialogue generation method for application software configuration of Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2747425C2 (en) | Real-time answer system to questions from different fields of knowledge | |
| Yu et al. | FlowSense: A natural language interface for visual data exploration within a dataflow system | |
| Weber et al. | Textual case-based reasoning | |
| CN112749263A (en) | Multi-round answer generation system based on single question | |
| US11853362B2 (en) | Using a multi-task-trained neural network to guide interaction with a query-processing system via useful suggestions | |
| CN109446306A (en) | Task-driven multi-turn dialogue-based intelligent question and answer method | |
| US20220343082A1 (en) | System and method for ensemble question answering | |
| CN114003709A (en) | Intelligent question-answering system and method based on question matching | |
| CN117493529B (en) | Anthropomorphic dialogue method and device based on natural language model and electronic equipment | |
| CN117370580A (en) | Knowledge-graph-based large language model enhanced dual-carbon field service method | |
| CN112507089A (en) | Intelligent question-answering engine based on knowledge graph and implementation method thereof | |
| CN116244344A (en) | Retrieval method and device based on user requirements and electronic equipment | |
| CN115982337A (en) | Intention identification method for transaction handling in online conversation | |
| US20230153522A1 (en) | Image captioning | |
| Yi et al. | A Survey on Recent Advances in LLM-Based Multi-turn Dialogue Systems | |
| CN118093829A (en) | Legal consultation method based on knowledge graph enhancement | |
| CN117438047A (en) | Psychological consultation model training and psychological consultation processing method and device and electronic equipment | |
| Arbaaeen et al. | Natural language processing based question answering techniques: A survey | |
| Trummer | Data Vocalization with CiceroDB. | |
| Melo et al. | Semantic web search through natural language dialogues | |
| Dai et al. | QAM: question answering system based on knowledge graph in the military | |
| Melo et al. | Using a dialogue manager to improve semantic web search | |
| CN118070925B (en) | Model training method, device, electronic equipment, storage medium and program product | |
| OsmanHamoud | A Question Answering System Design about the Holy Quran | |
| CN118070893B (en) | Information feedback method, device and computer program product based on large language model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210504 |