CN115795018A - Multi-strategy intelligent searching question-answering method and system for power grid field - Google Patents
Multi-strategy intelligent searching question-answering method and system for power grid field Download PDFInfo
- Publication number
- CN115795018A CN115795018A CN202310102066.0A CN202310102066A CN115795018A CN 115795018 A CN115795018 A CN 115795018A CN 202310102066 A CN202310102066 A CN 202310102066A CN 115795018 A CN115795018 A CN 115795018A
- Authority
- CN
- China
- Prior art keywords
- question
- user
- vector
- faq
- answer
- 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.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a multi-strategy intelligent search question and answer method and a multi-strategy intelligent search question and answer system for the field of power grids, which comprise the following steps: semantic feature extraction is carried out on the obtained user question and FAQ question-answer pair library, the user question is converted into a vector form, and semantic feature vectors are extracted from each FAQ question-answer pair; constructing a vector index for each question-answer pair; according to a vector form of the index corresponding to the user question, vector retrieval is carried out by using an FAISS, L2 normalization processing is carried out on a user question vector and an FAQ question-answer pair library vector index matrix, through cosine similarity calculation, the obtained scores are used as similarity bases, N vectors with the highest similarity to the user question are obtained, and N question-answer pairs with the highest similarity to the user question are obtained; and sequencing the N question-answer pairs with the highest similarity obtained based on the recall stage, and returning K answers with the highest similarity.
Description
Technical Field
The invention relates to the technical field of natural language processing, in particular to a multi-strategy intelligent question and answer searching method and system for the field of power grids.
Background
China is a large power utilization country, needs professional power technical workers to support power production and maintenance, is complex in field operation, variable in condition and high in danger coefficient, and needs to accurately judge problems and quickly solve the problems by maintenance personnel. For the relevant knowledge content of the professional power grid, the traditional question answering system cannot meet the existing requirements.
The FAQ question-answering system is an artificial intelligence product combining an information retrieval technology and a natural language processing technology. A common FAQ question-answering system generally requires the following three steps from the question-answering by a user to the answer obtaining by the user: question analysis, information retrieval and answer generation. Aiming at the problem provided by the user in the natural language form, the problem analysis module mainly realizes the pretreatment of the problem, generally carries out word segmentation, word stop removal and part of speech tagging on the problem, and simultaneously obtains the keyword of the problem so as to obtain the problem intention of the user. The information retrieval module mainly retrieves the most similar questions to the user questions from the given FAQ data set through similarity calculation. The answer generation module is mainly used for sequencing the retrieved question-answer pairs and returning the answer with the highest user similarity to the user.
In an FAQ question-answering system, text similarity calculation is a key technology in an information retrieval stage, and the current text similarity calculation methods generally comprise the following types:
(1) The method is based on a literal matching method, and mainly carries out text matching from the morphology, and typical methods are like a Jaacard distance, a longest common substring method, a BM25 algorithm and the like. In general, the same word may have different meanings in different contexts, and different words may have the same meaning in some scenarios, so that the literal matching method has a large limitation in semantic similarity.
(2) The method based on statistics mainly converts text into vector form according to corpus so that computer can execute corresponding operation to obtain words with similar context. Typical methods are Vector Space Model (VSM), neural network Model, etc. For example, in the DSSM network designed and proposed by Huang et al, a DNN network is used to map question and question-answer pairs into place semantic vectors, and then the cosine distance between the semantic vectors is calculated in a low-dimensional vector space, and the similarity is obtained to implement a text matching task, but the effect on the context information and the word order information of the text is not good enough.
(3) The method based on the rules needs to manually construct a knowledge base with certain rules and decompose the feature words in the text into concepts according to the defined rules, so that the similarity calculation between the feature words is converted into the similarity calculation between the concepts. The Explicit Semantic Analysis (ESA) proposed by Gabrilovich et al designs calculates the similarity between the text features by calculating the distance between the concept vectors to obtain a good effect, but it needs to consume a large amount of human resources in the construction of the rule part.
In an actual application scenario, namely in the industry, there is a high requirement for the matching speed of the FAQ question-answering system, the system is usually required to be capable of interacting with a user in real time, the problem form proposed by the user is complicated, the keyword omission and indication phenomena exist in the problem, and the spoken problem form brings a certain obstacle to the solution of the user problem to the computing mechanism. In addition, as events progress and product versions are iterated, the FAQ question-answer library needs to be updated incrementally at regular intervals, which also has a high requirement on the robustness of the FAQ question-answer system. Secondly, the academic community has few researches on the FAQ question-answering system, focuses on the single-point technology aspects such as similarity calculation and the like, cannot meet the requirements of practical application scenarios, and brings certain challenges to the overall research and application realization of the FAQ question-answering system.
Disclosure of Invention
The invention provides a multi-strategy intelligent question and answer searching method and system for the power grid field, aiming at solving the problem that the conventional question and answer function in the prior art can not accurately search the knowledge content in the power grid professional field, so that the problem that the conventional question and answer function can not accurately search the knowledge content in the power grid professional field at the semantic level is solved.
In order to realize the purpose of the invention, the technical scheme is as follows:
a multi-strategy intelligent searching question-answering method facing to the field of power grids comprises the following steps:
acquiring a user question;
meanwhile, semantic feature extraction is carried out on the user questions and an FAQ question-answer pair library formed by knowledge contents in the power grid professional field, the user questions are converted into a vector form, and semantic feature vectors are extracted from each FAQ question-answer pair;
constructing a vector index of each question-answer pair for the semantic feature vector of the FAQ question-answer pair by using an FAISS vector search engine, thereby obtaining a vector index matrix of an FAQ question-answer pair library;
according to a vector form of each question-answer pair corresponding to the user question, vector retrieval is carried out by using an FAISS, the user question vector and a constructed FAQ question-answer pair library vector index matrix are subjected to L2 normalization processing, then, through cosine similarity calculation, the obtained scores are used as similarity bases, N vectors with the highest similarity to the user question are obtained, and N question-answer pairs with the highest similarity to the user question are retrieved from an FAQ question-answer pair library;
sorting the N question-answer pairs with the highest similarity obtained based on the recall stage, and returning K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
Preferably, after the user question is obtained, whether the user question is an empty character string is judged; if the character string is empty, prompting the user to input again; and if the user question is not an empty character string, converting the user question into a vector form.
Preferably, a deep learning model is used to extract semantic feature vectors for each question-answer pair of the FAQ question-answer pair library.
Preferably, the index of each question-answer pair is constructed using inverted product quantization index pairs in the FAISS vector search engine.
Preferably, if no standard question similar to the user question is found, the system goes to a KBQA module, and the question asked by the user is processed by using the atlas question-answer service;
the KBQA module searches and matches whether similar standard problems exist in an FAQ problem-answer pair library, and if so, corresponding answers are returned; if the standard problem does not exist, adding the problem into an FAQ knowledge base and manually giving an answer; meanwhile, the problem of dissatisfaction fed back by the user is also added into an FAQ knowledge base and answers are given manually; and taking the manually given answers as training data samples of the next iteration for reconstructing the FAQ question-answer pair library.
Preferably, the deep learning model adopts a sensor-transform framework, and completes construction of FAQ question-answer to a library and extraction of semantic feature vectors of user questions by using a discrete-base-multilinual-case model in a pre-training model library with an open source of Hugging Face; the distiluse-base-multilingual-cast model structure is a double-tower model structure; the training process is that sentence pairs are respectively input into two sub BERT models, sentence representations u and v are obtained through a pooling strategy, the sentence representations u and v are spliced with | u-v |, the similar probability is obtained through a Softmax function, a partition-base-multilingual-coded model is obtained by minimizing cross entropy loss, and meanwhile the weight W is updated.
A multi-strategy intelligent search question-answering system facing the power grid field comprises an FAQ question-answer pair library, a semantic feature extraction module, a vector library index construction module, a semantic information retrieval module and a question matching module;
the FAQ question-answer pair library consists of knowledge contents in the professional field of the power grid;
the semantic feature extraction module is used for extracting semantic features of the user questions and converting the user questions into a vector form; meanwhile, extracting semantic features of an FAQ question-answer pair library, and extracting semantic feature vectors of each FAQ question-answer pair;
the vector library index construction module is used for constructing a vector index of each question-answer pair by using a FAISS vector search engine for semantic feature vectors of FAQ question-answer pairs, so as to obtain a vector index matrix of an FAQ question-answer pair library;
the semantic information retrieval module is used for carrying out vector retrieval by using an FAISS according to a vector form corresponding to the index of each question-answer pair and a user question, carrying out L2 normalization processing on a user question vector and a constructed FAQ question-answer pair library vector index matrix, then carrying out cosine similarity calculation to obtain scores as a similarity basis, obtaining N vectors with the highest similarity to the user question and obtaining N question-answer pairs with the highest similarity to the user question from an FAQ question-answer pair library;
the question matching module sorts the N question-answer pairs with the highest similarity obtained based on the recall stage and returns K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
Preferably, the semantic feature extraction module extracts feature vectors by using a deep learning model.
A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the grid-domain oriented multi-policy intelligent search question-answering method when executing the computer program.
A computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the grid-domain-oriented multi-policy intelligent search question-answering method.
The invention has the following beneficial effects:
according to the invention, the vector characteristics of each question sentence are extracted by adopting a semantic vector retrieval mode, and the most accurate and most similar question-answer pairs matched in the FAQ question-answer pair library collected from the frequently asked questions and the standard answers of the user under the actual condition are obtained through question vector calculation, so that the search accuracy is improved. Meanwhile, the vectors are processed by adopting a vector search engine in the vector matching process, so that the query and sequencing of massive similar texts can be realized very quickly, the accuracy and efficiency of problem reply are improved aiming at the repeatability problem provided by a user, the answers of common problems are returned to the user quickly, the problem that the system cannot reply can be solved by manual customer service, the working efficiency of personnel is improved, and meanwhile, the expenditure of labor resources and financial resources of enterprises is greatly saved.
Drawings
Fig. 1 is a flowchart of a multi-strategy intelligent question and answer searching method for the power grid field according to the present invention.
FIG. 2 is a schematic block diagram of a multi-strategy intelligent search question-answering method for the power grid field according to the present invention.
Fig. 3 is a diagram of a double tower model architecture.
Fig. 4 is a diagram of an FAQ question-answer pair generation example.
FIG. 5 is a schematic diagram of the multi-strategy intelligent search question-answering system of the present invention.
Detailed Description
The invention is described in detail below with reference to the drawings and the detailed description.
Example 1
As shown in fig. 1 and 2, a multi-strategy intelligent question and answer searching method for the power grid field includes the following steps:
acquiring a user question;
meanwhile, semantic feature extraction is carried out on the user questions and an FAQ question-answer pair library formed by knowledge contents of the professional field of the power grid, the user questions are converted into a vector form, and semantic feature vectors are extracted from each FAQ question-answer pair;
constructing a vector index of each question-answer pair for the semantic feature vector of the FAQ question-answer pair by using an FAISS vector search engine, thereby obtaining a vector index matrix of an FAQ question-answer pair library;
according to the vector form of the index of each question-answer pair and the user question, vector retrieval is carried out by using an FAISS, the user question vector and the constructed vector index matrix of the FAQ question-answer pair library are subjected to L2 normalization processing, then through cosine similarity calculation, the obtained scores are used as the similarity basis, N vectors with the highest similarity to the user question are obtained, and N question-answer pairs with the highest similarity to the user question are retrieved from the FAQ question-answer pair library;
sorting the N question-answer pairs with the highest similarity obtained based on the recall scoring stage, and returning K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
The embodiment can quickly search the feature vectors of the user according to the vector search technology aiming at the problems provided by the user.
In this embodiment, the step of retrieving N question-answer pairs with the highest similarity to the user question from the FAQ question-answer pair library belongs to the recall scoring stage, and the cos cosine distance is calculated by using the FAISS as the similarity score. The K answers with the highest returned similarity belong to the sorting stage, a certain number of returns can be performed according to the requirements of the user, and K is the topK of N. The set of vectors constructed by the present embodiment using the FAISS is stored in the matrix. The FAISS uses only a 32-bit floating-point matrix, where the columns of the matrix represent the characteristics of the vector and the rows represent the number of vector samples.
In a specific embodiment, after the user question is obtained, whether the user question is an empty character string is judged; if the character string is empty, prompting the user to input again; and if the user question is not an empty character string, converting the user question into a vector form.
In a particular embodiment, a deep learning model is used to extract semantic feature vectors for each question-answer pair of the FAQ question-answer pair library. The deep learning model adopts a sensor transformations model.
In a specific embodiment, an inverted product quantization index pair in an FAISS vector search engine is used for constructing an index of each question-answer pair for FAQ question-answer pairs, so that quick retrieval of user question expression vectors is realized, and the question-answer efficiency of an FAQ question-answer system is improved.
In a specific embodiment, if a standard question similar to the user question is not found, the system goes to a KBQA module, and the question asked by the user is processed by using the atlas question-answer service;
the KBQA module searches and matches whether similar standard problems exist in an FAQ problem-answer pair library, and if so, returns a corresponding answer; if the standard problem does not exist, adding the problem into an FAQ knowledge base and manually giving an answer; meanwhile, unsatisfactory answers fed back by the user are also added into the FAQ knowledge base and are given manually; and taking the manually given answers as training data samples of the next iteration for reconstructing the FAQ question-answer pair library.
In this embodiment, the iteration time is set to be one iteration per week, and the FAQ question-answer pair library is reconstructed to further optimize the intelligent question-answer system.
The KBQA is an atlas question-answer module, is an independent service, and calls the KBQA module to answer the question of the user when no answer is obtained from the FAQ question-answer pair library.
It should be noted that, as shown in fig. 4, in order to further improve the accuracy, after returning the answer, the satisfaction degree of the user on the answer is also marked, such as satisfaction or dissatisfaction; the method is favorable for submitting question-answer accuracy, can ensure to return answers expected by the user, and improves the satisfaction degree of the user.
In a specific embodiment, as shown in fig. 3, the deep learning model adopts a sensor-transform framework, and completes the construction of the FAQ question-answer pair library and the extraction of the semantic feature vector of the user question by using a discrete-base-multilingual-cast model in a Hugging Face open-source pre-training model library; the distiluse-base-multilingual-cast model structure is a double-tower model structure; the training process is that sentence pairs are respectively input into two sub BERT models, sentence representations u and v are obtained through a pooling strategy, the sentence representations u and v are spliced with | u-v |, the similar probability is obtained through a Softmax function, a partition-base-multilingual-coded model is obtained by minimizing cross entropy loss, and meanwhile the weight W is updated.
The discrete-base-multilingualal-case model is a pre-training model trained by Hugging Face through large-scale corpora.
In the embodiment, by adopting a semantic vector retrieval mode and utilizing a vectorization technology, a pre-training model trained by a large number of FAQ question-answer pair libraries is used to extract the vector characteristics of each sentence, and the semantic distance between a question and a candidate question in the FAQ question-answer pair libraries is obtained through vector calculation, so that the problem that the existing traditional question-answer function cannot accurately search the semantic level of knowledge content in the professional field of the power grid is solved. And the problem of low accuracy of single question-answer retrieval is solved through the combination of the FAQ module and the KBQA module.
Example 2
As shown in fig. 5, a multi-strategy intelligent search question-answering system for the power grid field includes an FAQ question-answer pair library, a semantic feature extraction module, a vector library index construction module, a semantic information retrieval module, and a question matching module;
the FAQ question-answer pair library consists of knowledge contents in the professional field of the power grid;
the semantic feature extraction module is used for extracting semantic features of the user questions and converting the user questions into a vector form; meanwhile, extracting semantic features of an FAQ question-answer pair library, and extracting semantic feature vectors of each FAQ question-answer pair;
the vector library index constructing module is used for constructing a vector index of each question-answer pair for the semantic feature vector of the FAQ question-answer pair by using an FAISS vector search engine, so as to obtain a vector index matrix of an FAQ question-answer pair library;
the semantic information retrieval module is used for carrying out vector retrieval by using an FAISS according to a vector form corresponding to the index of each question-answer pair and a user question, carrying out L2 normalization processing on a user question vector and a constructed FAQ question-answer pair library vector index matrix, then carrying out cosine similarity calculation to obtain scores as a similarity basis, obtaining N vectors with the highest similarity to the user question and obtaining N question-answer pairs with the highest similarity to the user question from an FAQ question-answer pair library;
the question matching module sorts the N question-answer pairs with the highest similarity obtained based on the recall stage and returns K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
In this embodiment, the semantic feature extraction module extracts feature vectors by using a deep learning model.
Example 3
A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the grid-domain oriented multi-policy intelligent search question-answering method according to embodiment 1.
Where the memory and processor are connected by a bus, the bus may comprise any number of interconnected buses and bridges, the buses connecting together one or more of the various circuits of the processor and the memory. The bus may also connect various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. The data processed by the processor is transmitted over a wireless medium via an antenna, which further receives the data and transmits the data to the processor.
Example 4
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the grid-domain-oriented multi-policy intelligent search question-answering method according to embodiment 1.
That is, as can be understood by those skilled in the art, all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, or the like) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A multi-strategy intelligent searching question-answering method facing the power grid field is characterized in that: the method comprises the following steps:
acquiring a user question;
meanwhile, semantic feature extraction is carried out on the user questions and an FAQ question-answer pair library formed by knowledge contents in the power grid professional field, the user questions are converted into a vector form, and semantic feature vectors are extracted from each FAQ question-answer pair;
constructing a vector index of each question-answer pair for the semantic feature vector of the FAQ question-answer pair by using an FAISS vector search engine, thereby obtaining a vector index matrix of an FAQ question-answer pair library;
according to a vector form of each question-answer pair corresponding to the user question, vector retrieval is carried out by using an FAISS, the user question vector and a constructed FAQ question-answer pair library vector index matrix are subjected to L2 normalization processing, then, through cosine similarity calculation, the obtained scores are used as similarity bases, N vectors with the highest similarity to the user question are obtained, and N question-answer pairs with the highest similarity to the user question are retrieved from an FAQ question-answer pair library;
sorting the N question-answer pairs with the highest similarity obtained based on the recall stage, and returning K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
2. The power grid field-oriented multi-strategy intelligent search question-answering method according to claim 1, characterized in that: after the user problem is obtained, judging whether the user problem is an empty character string or not; if the character string is empty, prompting the user to input again; and if the user question is not an empty character string, converting the user question into a vector form.
3. The power grid field-oriented multi-strategy intelligent search question-answering method according to claim 1, characterized in that: semantic feature vectors are extracted for each question-answer pair of the FAQ question-answer pair library using a deep learning model.
4. The power grid field-oriented multi-strategy intelligent search question-answering method according to claim 1, characterized in that: the index of each question-answer pair is constructed using inverted product quantization index pairs in the FAISS vector search engine.
5. The power grid field-oriented multi-strategy intelligent search question-answering method according to claim 1, characterized in that: if the standard question similar to the user question is not found, the system is switched to a KBQA module, and the question asked by the user is processed by using the atlas question-answer service;
the KBQA module searches and matches whether similar standard problems exist in an FAQ problem-answer pair library, and if so, corresponding answers are returned; if the standard problem does not exist, adding the problem into an FAQ knowledge base and manually giving an answer; meanwhile, the problem of dissatisfaction fed back by the user is also added into an FAQ knowledge base and answers are given manually; and taking the manually given answers as training data samples of the next iteration for reconstructing the FAQ question-answer pair library.
6. The power grid field-oriented multi-strategy intelligent search question-answering method according to claim 3, characterized in that: the deep learning model adopts a sensor-transform framework, and completes the construction of an FAQ question-answer pair library and the extraction of semantic feature vectors of user questions by using a discrete-base-multilingual-case model in a pre-training model library with an open source of Hugging Face; the method comprises the following steps of firstly inputting sentence pairs into two sub BERT models respectively, then obtaining sentence representations u and v through a pooling strategy, splicing the sentence representations u and v and | u-v |, finally obtaining similar probability through a Softmax function, and simultaneously updating the weight W through minimizing cross entropy loss to obtain the distilution-base-multilingual-cased model.
7. A multi-strategy intelligent search question-answering system oriented to the field of power grids is characterized in that: the system comprises an FAQ question-answer pair library, a semantic feature extraction module, a vector library index construction module, a semantic information retrieval module and a question matching module;
the FAQ question-answer pair library consists of knowledge contents in the professional field of the power grid;
the semantic feature extraction module is used for extracting semantic features of the user questions and converting the user questions into a vector form; meanwhile, extracting semantic features of an FAQ question-answer pair library, and extracting semantic feature vectors of each FAQ question-answer pair;
the vector library index constructing module is used for constructing a vector index of each question-answer pair for the semantic feature vector of the FAQ question-answer pair by using an FAISS vector search engine, so as to obtain a vector index matrix of an FAQ question-answer pair library;
the semantic information retrieval module is used for carrying out vector retrieval by using an FAISS according to a vector form corresponding to the index of each question-answer pair and a user question, carrying out L2 normalization processing on a user question vector and a constructed FAQ question-answer pair library vector index matrix, then carrying out cosine similarity calculation to obtain scores as a similarity basis, obtaining N vectors with the highest similarity to the user question and obtaining N question-answer pairs with the highest similarity to the user question from an FAQ question-answer pair library;
the question matching module sorts the N question-answer pairs with the highest similarity obtained based on the recall stage and returns K answers with the highest similarity; n, K are all positive integers greater than 0, and K is less than or equal to N.
8. The power grid domain-oriented multi-strategy intelligent search question-answering system according to claim 7, wherein: the semantic feature extraction module adopts a deep learning model to extract feature vectors.
9. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the grid-domain oriented multi-strategy intelligent search question-answering method according to any one of claims 1 to 6.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that: the computer program, when executed by a processor, implements the steps of the grid domain oriented multi-strategy intelligent search question-answering method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310102066.0A CN115795018B (en) | 2023-02-13 | 2023-02-13 | Multi-strategy intelligent search question-answering method and system for power grid field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310102066.0A CN115795018B (en) | 2023-02-13 | 2023-02-13 | Multi-strategy intelligent search question-answering method and system for power grid field |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115795018A true CN115795018A (en) | 2023-03-14 |
| CN115795018B CN115795018B (en) | 2023-05-09 |
Family
ID=85430935
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310102066.0A Active CN115795018B (en) | 2023-02-13 | 2023-02-13 | Multi-strategy intelligent search question-answering method and system for power grid field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115795018B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111046155A (en) * | 2019-11-27 | 2020-04-21 | 中博信息技术研究院有限公司 | Semantic similarity calculation method based on FSM multi-turn question answering |
| CN112559723A (en) * | 2020-12-28 | 2021-03-26 | 广东国粒教育技术有限公司 | FAQ search type question-answer construction method and system based on deep learning |
| CN109271505B (en) * | 2018-11-12 | 2021-04-30 | 深圳智能思创科技有限公司 | Question-answering system implementation method based on question-answer pairs |
| CN108804521B (en) * | 2018-04-27 | 2021-05-14 | 南京柯基数据科技有限公司 | Knowledge graph-based question-answering method and agricultural encyclopedia question-answering system |
| US11210306B2 (en) * | 2019-11-08 | 2021-12-28 | PolyAI Limited | Dialogue system, a method of obtaining a response from a dialogue system, and a method of training a dialogue system |
| CN112507097B (en) * | 2020-12-17 | 2022-11-18 | 神思电子技术股份有限公司 | Method for improving generalization capability of question-answering system |
| CN115470338A (en) * | 2022-10-27 | 2022-12-13 | 之江实验室 | Multi-scene intelligent question and answer method and system based on multi-way recall |
| CN115510814A (en) * | 2022-11-09 | 2022-12-23 | 东南大学 | Chapter-level complex problem generation method based on double planning |
| WO2023273170A1 (en) * | 2021-06-30 | 2023-01-05 | 同济人工智能研究院(苏州)有限公司 | Welcoming robot conversation method |
-
2023
- 2023-02-13 CN CN202310102066.0A patent/CN115795018B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108804521B (en) * | 2018-04-27 | 2021-05-14 | 南京柯基数据科技有限公司 | Knowledge graph-based question-answering method and agricultural encyclopedia question-answering system |
| CN109271505B (en) * | 2018-11-12 | 2021-04-30 | 深圳智能思创科技有限公司 | Question-answering system implementation method based on question-answer pairs |
| US11210306B2 (en) * | 2019-11-08 | 2021-12-28 | PolyAI Limited | Dialogue system, a method of obtaining a response from a dialogue system, and a method of training a dialogue system |
| CN111046155A (en) * | 2019-11-27 | 2020-04-21 | 中博信息技术研究院有限公司 | Semantic similarity calculation method based on FSM multi-turn question answering |
| CN112507097B (en) * | 2020-12-17 | 2022-11-18 | 神思电子技术股份有限公司 | Method for improving generalization capability of question-answering system |
| CN112559723A (en) * | 2020-12-28 | 2021-03-26 | 广东国粒教育技术有限公司 | FAQ search type question-answer construction method and system based on deep learning |
| WO2023273170A1 (en) * | 2021-06-30 | 2023-01-05 | 同济人工智能研究院(苏州)有限公司 | Welcoming robot conversation method |
| CN115470338A (en) * | 2022-10-27 | 2022-12-13 | 之江实验室 | Multi-scene intelligent question and answer method and system based on multi-way recall |
| CN115510814A (en) * | 2022-11-09 | 2022-12-23 | 东南大学 | Chapter-level complex problem generation method based on double planning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115795018B (en) | 2023-05-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117033608B (en) | Knowledge graph generation type question-answering method and system based on large language model | |
| CN109271505B (en) | Question-answering system implementation method based on question-answer pairs | |
| CN106874378B (en) | Method for constructing knowledge graph based on entity extraction and relation mining of rule model | |
| CN107818164A (en) | A kind of intelligent answer method and its system | |
| CN106951558B (en) | Data processing method of tax intelligent consultation platform based on deep search | |
| CN105095444A (en) | Information acquisition method and device | |
| CN112989004B (en) | Query graph ordering method and system for knowledge graph question answering | |
| CN109947952B (en) | Retrieval method, device, equipment and storage medium based on English knowledge graph | |
| CN111966810A (en) | Question-answer pair ordering method for question-answer system | |
| CN112463944A (en) | Retrieval type intelligent question-answering method and device based on multi-model fusion | |
| CN113947084A (en) | Question-answer knowledge retrieval method, device and equipment based on graph embedding | |
| CN115858750A (en) | Power grid technical standard intelligent question-answering method and system based on natural language processing | |
| CN111881264B (en) | Method and electronic equipment for searching long text in question-answering task in open field | |
| CN110390050B (en) | Software development question-answer information automatic acquisition method based on deep semantic understanding | |
| CN110851584B (en) | Legal provision accurate recommendation system and method | |
| CN116775846A (en) | Domain knowledge question and answer method, system, equipment and medium | |
| CN116401344A (en) | Method and device for searching table according to question | |
| CN113468311B (en) | Knowledge graph-based complex question and answer method, device and storage medium | |
| CN115795018B (en) | Multi-strategy intelligent search question-answering method and system for power grid field | |
| CN114997331A (en) | Small sample relation classification method and system based on metric learning | |
| CN114417863A (en) | Word weight generation model training method and device and word weight generation method and device | |
| CN114579706A (en) | Automatic subjective question evaluation method based on BERT neural network and multitask learning | |
| CN114003706A (en) | Keyword combination generation model training method and device | |
| CN108241650B (en) | Training method and device for training classification standard | |
| CN116089589B (en) | Question generation method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |