Intelligent question and answer method, device and equipment and storage medium

Abstract

The invention discloses an intelligent question and answer method and device, equipment and a storage medium. The method comprises the following steps: receiving initial problem data input by a user, and determining related document data associated with the initial problem data based on the initial problem data and a pre-established document database; generating target problem data based on the initial problem data and the related document data; and inputting the target question data into an intelligent question and answer model, and determining target result data based on a model output result. And the accuracy of intelligent question answering is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent question answering, and in particular to an intelligent question answering method, device, equipment and storage medium. BACKGROUND

[0002] The emergence of large language models has changed the way we interact with information, demonstrating amazing versatility and intelligence.

[0003] However, these models are not perfect and can produce misleading answers, rely on outdated information, be inefficient in handling specific knowledge, and have limitations in depth and reasoning ability in specialized fields. In practical applications, to cope with changing environments and protect data privacy, data needs to be updated and the transparency and traceability of generated content need to be ensured. SUMMARY

[0004] The present application provides an intelligent question answering method, device, equipment and storage medium to solve the problem of low efficiency in handling specific knowledge and limitations in depth and reasoning ability in specialized fields of existing large language models in the question answering field.

[0005] According to an aspect of the present application, an intelligent question answering method is provided, the method comprising:

[0006] receiving initial question data input by a user, determining relevant document data associated with the initial question data based on the initial question data and a pre-established document database;

[0007] generating target question data based on the initial question data and the relevant document data;

[0008] inputting the target question data into an intelligent question answering model, and determining target result data based on the model output result.

[0009] According to another aspect of the present application, an intelligent question answering device is provided, the device comprising:

[0010] an initial question receiving module configured to receive initial question data input by a user, and determine relevant document data associated with the initial question data based on the initial question data and a pre-established document database;

[0011] a target question generating module configured to generate target question data based on the initial question data and the relevant document data;

[0012] an intelligent question answering module configured to input the target question data into an intelligent question answering model, and determine target result data based on the model output result.

[0013] According to another aspect of the present application, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected with the at least one processor; wherein,

[0016] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the intelligent question-answering method according to any one of the embodiments of the present application.

[0017] According to another aspect of the present application, a computer readable storage medium is provided, which stores computer instructions for enabling a processor to implement the intelligent question-answering method according to any one of the embodiments of the present application when executed.

[0018] The technical solution of the embodiments of the present application receives initial question data input by a user, determines relevant document data associated with the initial question data based on the initial question data and a pre-established document database; implements semantic-level retrieval through the pre-established document database, avoids the limitation of keyword matching, and ensures that the returned documents are relevant to the deep semantics of the question; then, generates target question data based on the initial question data and the relevant document data; converts fuzzy queries into structured queries in combination with the context information of the relevant documents; inputs the target question data into an intelligent question-answering model, and determines target result data based on the output result of the model. The problem of low accuracy of intelligent question-answering is solved, and the beneficial effect of improving the accuracy of intelligent question-answering is achieved.

[0019] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0021] Figure 1 is a flowchart of an intelligent question-answering method according to an embodiment of the present application;

[0022] Figure 2 is a flowchart of an intelligent question-answering method according to an embodiment of the present application;

[0023] Figure 3is a structural schematic diagram of an intelligent question answering device according to embodiment three of the present application;

[0024] Figure 4 is a structural schematic diagram of an electronic device implementing an intelligent question answering method of the present application. DETAILED DESCRIPTION

[0025] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] Embodiment one

[0028] Figure 1 A flowchart of an intelligent question answering method is provided for embodiment one of the present application. The present embodiment can be applicable to the case of intelligent question answering. The method can be executed by an intelligent question answering device, which can be implemented in the form of hardware and / or software. The intelligent question answering device can be configured in an electronic device. As shown in the figure, the method comprises: Figure 1

[0029] S110, receiving initial question data input by a user, and determining relevant document data associated with the initial question data based on the initial question data and based on the initial question data.

[0030] The initial question data can be understood as the question or query content initially input by the user. The relevant document data can be understood as the document information closely associated with the initial question data of the user found from the document database.

[0031] ​Specifically, the question or query content proposed by the user is acquired, and relevant document data associated with the initial question data is searched based on the initial question data.

[0032] Optionally, before determining the relevant document data associated with the initial question data based on the initial question data and the pre-established document database, the method further includes: acquiring a preset number of sample data, and extracting document data in the sample data respectively; for each document data, encoding text in the document data into a vector through an encoding model, and constructing a document database based on a plurality of vectors, wherein the document database stores indexes between a plurality of vectors and a plurality of document data.

[0033] The sample data can be understood as an initial document set for constructing the document database. The encoding model is a model for converting text data into a vector. The vector can be understood as a numerical representation obtained by converting the text data through the encoding model. The index can be understood as a mapping relationship between the document vector and the document data.

[0034] Specifically, a certain number of documents are collected as the basis for constructing the database, the pure text content is extracted from the sample data to form a document set to be encoded. The text of each document is converted into a numerical vector through a TF-IDF model, and the document identifier, the vector and the text content are integrated into the database to establish a mapping relationship between the vector and the document.

[0035] Exemplarily, considering that the data in the data lake includes structured (SQL, csv, excel, etc.), semi-structured (json, etc.) and unstructured (PDF, word, html, etc.) data, in order to facilitate the later query index, it is necessary to uniformly process the above different data into the same format. In this step, the application first parses different data to extract the text content therein; secondly, the application divides the extracted text content into independent units according to the number of sentences. Each unit contains 15 sentences. In order to retain better context semantic information, the application sets an overlapping sentence between different units. If the text content is less than 15 sentences, the text content is used as an independent unit.

[0036] Exemplarily, the embodiment encodes the parsed text data into vectors by an encoding model (for example, BGE v1.5 encoding model). The BGE v1.5 encoding model adopts the same Transformer architecture as the BERT model, and adopts the vector corresponding to the [CLS] token as the vector representation of the input text; the model is trained on a large-scale Chinese data set containing 100M text pairs, and has good general text encoding capability. Specifically, the text content in the independent unit is input into the BGE v1.5-Large encoding model; after forward inference, the vector corresponding to the [CLS] token in the output is saved to the locally built vector database. For data content in the banking field, more proper nouns are often contained, so in the implementation process, the model can be slightly tuned for the specific nouns in the financial industry, so that the vectorization process is more suitable for the construction of the bank field index vector library.

[0037] Exemplarily, the embodiment builds a vector database based on FAISS on the local server, which can save the index between vectors and texts. The FAISS framework can realize efficient similar vector query through the clustering algorithm of inverted index IVF and product quantization PQ.

[0038] S120, generating target question data based on the initial question data and the related document data.

[0039] The target question data can be understood as a new question generated based on the initial question data and the related document data.

[0040] Specifically, the initial question data and the related document data are combined to generate target question data.

[0041] Optionally, the generating target question data based on the initial question data and the related document data comprises:

[0042] Combining the initial question data and the related document data into target question data based on a preset question template.

[0043] The preset question template can be understood as a pre-defined question structure for guiding the generation of target questions. It can be pre-set according to the demand, and the embodiment does not limit it.

[0044] Specifically, the core words are extracted from the initial question and the related document, and the extracted keywords are filled into the placeholders of the preset template to form a complete question.

[0045] S130, inputting the target question data into an intelligent question and answer model, and determining target result data based on the model output result.

[0046] The intelligent question and answer model can be understood as a large language model for processing questions and generating answers. The target result data can be understood as a set of answers output by the intelligent question and answer model to the target question data.

[0047] Specifically, the intelligent question and answer is performed based on the initial question data and the retrieved associated document data to obtain more comprehensive target result data.

[0048] Optionally, the target result data includes first result data corresponding to the initial question data and second result data corresponding to the target question data.

[0049] The first result data is the result of the user question. The second result data is the result of the user question combined with the relevant document.

[0050] Specifically, the first result data and the second result data are combined to obtain the target result data.

[0051] Illustratively, the large language model of the bank private domain real-time knowledge management question and answer tool used by the present application is used. After 30 text contents related to the user question are retrieved, a suitable prompt word is designed, and the retrieved text contents are combined with the user question as context information. The prompt word is a text segment guiding the output of the large language model, and its structure is shown in the following table:

[0052]

[0053]

[0054] Table 1

[0055] The present application inputs the set prompt word and the initial question proposed by the user into the intelligent question and answer model. The model generates answers for the user question based on the information retrieved from the bank private domain database. Finally, the present application feeds back the answers to the user in the front-end page.

[0056] The technical scheme of the embodiment of the present application receives the initial question data input by the user, determines the relevant document data associated with the initial question data based on the initial question data and the pre-established document database, realizes semantic-level retrieval through the pre-built document database, avoids the limitation of keyword matching, ensures that the returned document is related to the deep semantic of the question, then generates target question data based on the initial question data and the relevant document data, converts the fuzzy query into a structured query by combining the context information of the relevant document, inputs the target question data into the intelligent question and answer model, and determines the target result data based on the model output result. The problem of low accuracy of intelligent question and answer is solved, and the beneficial effect of improving the accuracy of intelligent question and answer is achieved.

[0057] Embodiment Two

[0058] Figure 2 A flowchart of an intelligent question answering method provided for Embodiment Two of the present application, this embodiment is a further refinement of how to generate target question data based on the initial question data and the relevant document data in the above-mentioned embodiments. Optionally, the determining of the relevant document data associated with the initial question data based on the initial question data and the pre-established document database comprises: splitting the initial question data into a preset number of sub-question data, for each sub-question data, searching for associated document data in the pre-established document database based on the sub-question data; and determining the relevant document data associated with the initial question data based on at least one of the associated document data.

[0059] As shown in Figure 2 , the method comprises:

[0060] S210, receiving initial question data input by a user, splitting the initial question data into a preset number of sub-question data, for each sub-question data, searching for associated document data in a pre-established document database based on the sub-question data.

[0061] Among them, the sub-question data can be understood as the question after splitting the initial question.

[0062] Specifically, this embodiment extracts n sub-questions {Q1,...,Qn} suitable for retrieval from the user question Q with the help of a large language model, and uses the user question Q and the generated n sub-questions {Q,Q1,...,Qn} to perform content retrieval respectively, and then fuses the retrieved content as context information of the large language model.

[0063] For example, the present application sets n to 5, and the split sub-questions are shown in the following table:

[0064]

[0065] Table 2

[0066] Optionally, the searching for associated document data in the pre-established document database based on the sub-question data comprises: searching for associated document data in the pre-established document database based on a similarity dense retrieval method and the sub-question data.

[0067] Among them, the similarity dense retrieval method can be understood as a technology that realizes accurate matching by calculating the cosine similarity between text vectors.

[0068] Specifically, the sub-problems and the documents are standardized, the text is converted into vectors using TF-IDF, a dense vector representation is simulated, and the cosine similarity between the sub-problem vector and the document vector is calculated. For example, the similarity between "the core module of bank private domain knowledge management" and document 1 is 0.0, the top 3 associated documents are sorted and output according to the similarity.

[0069] Optionally, the associated document data is found in the pre-established document database based on the similarity dense retrieval method and the sub-problem data, including: determining the similarity index corresponding to each document in the document database and the sub-problem data based on the similarity dense retrieval method; determining the similarity index sorting data based on a plurality of similarity indexes, and determining the associated document data corresponding to the sub-problem data based on the similarity index sorting data.

[0070] The similarity index can be understood as the numerical value of the similarity score. The similarity index sorting data can be understood as the order data obtained by sorting the similarity from high to low.

[0071] Specifically, a pre-trained deep learning model (such as BERT, Sentence-BERT) is used to convert each document and sub-problem data into a low-dimensional dense vector. These vectors can capture the semantic information of the text. By calculating the cosine similarity (or dot product score) between the sub-problem vector and each document vector, a numerical similarity index is generated. The similarity indexes of all documents are sorted from high to low to generate similarity index sorting data. For example, the Top 5 document IDs and scores may be: [(doc2, 0.92), (doc0, 0.85), (doc1, 0.78)]. According to a pre-set threshold (such as similarity > 0.7) or a fixed number (such as Top 3), the high similarity documents are extracted from the sorting data as the final associated document data. For example, doc2 and doc0 are selected as the most relevant answer basis for the sub-problem.

[0072] For example, the similarity dense retrieval method based on L2 distance is used in the embodiment of the present application to query the relevant context information for each problem. For Q1, the similarity score set based on L2 distance is:

[0073] S1=||e1-b i ||2,i=1,…,N

[0074] Where e1 is the vector encoding of Q1, b iis the i-th vector in the document database, and N is the total number of vectors in the document database. After obtaining the similarity scores, the embodiment ranks all the scores in the set S1 and takes the top 5 vectors with the highest similarity scores as the output. The same steps are performed for other sub-questions. Finally, the embodiment retrieves a total of 30 vectors related to the user question and takes the corresponding text content as the context information of the large language model.

[0075] The technical solution of the embodiment of the application reduces semantic ambiguity by decomposing a complex question into specific sub-questions, each of which focuses on a single topic.

[0076] S220, determining the relevant document data associated with the initial question data based on at least one of the associated document data.

[0077] Specifically, all the associated document data can be determined as the relevant document data associated with the initial question data. Alternatively, a preset number of associated document data can be selected as the relevant document data associated with the initial question data. For example, the top 3 associated document data in terms of relevance indicators or 3 associated document data randomly selected, which are not limited by the embodiment.

[0078] S230, generating target question data based on the initial question data and the relevant document data.

[0079] S240, inputting the target question data to an intelligent question and answer model and determining target result data based on the model output result.

[0080] The technical solution of the embodiment of the application decomposes the initial question data into a preset number of sub-question data, for each sub-question data, finds associated document data based on the sub-question data in a pre-established document database, and determines the relevant document data associated with the initial question data based on at least one of the associated document data. The complex question is decomposed into specific sub-questions, multiple sub-questions are retrieved in parallel, different dimensions of the initial question are covered, and a stereoscopic answer is formed after the associated documents are integrated, thereby improving the comprehensiveness of intelligent question and answer.

[0081] Embodiment Three

[0082] Figure 3 A structural schematic diagram of an intelligent question and answer device provided by the embodiment of the application is shown in FIG. 3. Figure 3 As shown in the figure, the device includes an initial question receiving module 310, a target question generating module 320, and an intelligent question and answer module 330.

[0083] The initial question receiving module 310 is configured to receive initial question data input by a user, and determine relevant document data associated with the initial question data based on the initial question data and a pre-established document database.

[0084] The technical scheme of the embodiment of the present application receives initial question data input by a user, and determines relevant document data associated with the initial question data based on the initial question data and a pre-established document database. Semantic-level retrieval is realized through the pre-established document database, the limitation of keyword matching is avoided, and the returned document is ensured to be related to the deep semantic of the question. Then, target question data is generated based on the initial question data and the relevant document data. Fuzzy query is converted into structured query by combining the context information of the relevant document. The target question data is input into an intelligent question answering model, and target result data is determined based on the output result of the model. The problem of low accuracy of intelligent question answering is solved, and the beneficial effect of improving the accuracy of intelligent question answering is achieved.

[0085] Optionally, the initial question receiving module comprises:

[0086] The question splitting unit is configured to split the initial question data into a preset number of sub-question data, and for each sub-question data, find associated document data in the pre-established document database based on the sub-question data.

[0087] The associated data determining unit is configured to determine the relevant document data associated with the initial question data based on at least one of the associated document data.

[0088] Optionally, the question splitting unit is specifically configured to:

[0089] Find the associated document data in the pre-established document database based on a similarity dense retrieval method and the sub-question data.

[0090] Optionally, the question splitting unit comprises:

[0091] The index determining sub-unit is configured to determine a similarity index corresponding to each document in the document database and the sub-question data based on the similarity dense retrieval method.

[0092] The index sorting sub-unit is configured to determine similarity index sorting data based on a plurality of the similarity indexes, and determine the associated document data corresponding to the sub-question data based on the similarity index sorting data.

[0093] Optionally, the device further comprises:

[0094] a document data extraction module configured to extract document data in the sample data respectively before determining the relevant document data associated with the initial question data based on the initial question data and the pre-established document database;

[0095] a database construction module configured to encode text in the document data into a vector by using an encoding model for each document data, and to construct a document database based on the plurality of vectors, wherein the document database stores indexes between the plurality of vectors and the plurality of document data.

[0096] Optionally, the target question generation module is specifically configured to:

[0097] combine the initial question data and the relevant document data into target question data based on a preset question template.

[0098] Optionally, the target result data includes first result data corresponding to the initial question data and second result data corresponding to the target question data.

[0099] The intelligent question and answer device provided by the embodiments of the present application can execute the intelligent question and answer method provided by any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0100] Embodiment Four

[0101] Figure 4 A structural schematic diagram of an electronic device 10 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present application described and / or claimed in this document.

[0102] As Figure 4As shown, the electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., communicatively connected to the at least one processor 11, where the memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0103] Various components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc., an output unit 17, such as various types of displays, a speaker, etc., a storage unit 18, such as a magnetic disk, an optical disk, etc., and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0104] The processor 11 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the method intelligent question answering.

[0105] In some embodiments, the method intelligent question answering can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded onto the RAM 13 and executed by the processor 11, one or more steps of the method intelligent question answering described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the method intelligent question answering by any other appropriate means, such as by means of firmware.

[0106] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0107] Computer programs used to implement the processes of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program, when executed, can cause instructions defined in the flow charts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package and partially on a remote machine or entirely on a remote machine or server.

[0108] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0109] To provide for interaction with a service acquirer, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the service acquirer, and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the service acquirer can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a service acquirer; for example, feedback provided to the service acquirer can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the service acquirer can be received in any form, including acoustic, speech, or tactile input.

[0110] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), blockchain network, and the Internet.

[0111] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0112] It should be understood that the steps shown in the above various forms of flow can be reordered, added, or deleted. For example, each step described in the present application can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of the present application can be achieved, and the present application does not limit herein.

[0113] The above detailed description does not limit the scope of the application. Various modifications, combinations, sub-combinations and alternatives can be made to the detailed description. Any modification, equivalent replacement and improvement etc. made within the spirit and principle of the application shall be included in the scope of the application.

Claims

1. An intelligent question answering method, characterized by, The method comprises: receiving initial question data input by a user, determining relevant document data associated with the initial question data based on the initial question data and a pre-established document database; generating target question data based on the initial question data and the relevant document data; inputting the target question data into an intelligent question-answering model, and determining target result data based on a model output result.

2. The method of claim 1, wherein, The generating of the target question data based on the initial question data and the relevant document data comprises: splitting the initial question data into a preset number of sub-question data, for each sub-question data, searching for associated document data in the pre-established document database based on the sub-question data; determining the relevant document data associated with the initial question data based on at least one of the associated document data.

3. The method of claim 2, wherein, The searching for the associated document data in the pre-established document database based on the sub-question data comprises: searching for the associated document data in the pre-established document database based on a similarity dense retrieval method and the sub-question data.

4. The method of claim 3, wherein, The searching for the associated document data in the pre-established document database based on the similarity dense retrieval method and the sub-question data comprises: determining, based on the similarity dense retrieval method, a similarity index corresponding to each document in the document database and the sub-question data; determining similarity index sorting data based on a plurality of the similarity indexes, and determining the associated document data corresponding to the sub-question data based on the similarity index sorting data.

5. The method of claim 1, wherein, Before the determining of the relevant document data associated with the initial question data based on the initial question data and the pre-established document database, the method further comprises: obtaining a preset number of sample data, and extracting document data in the sample data respectively; for each document data, encoding text in the document data into a vector through an encoding model, and constructing a document database based on a plurality of the vectors, wherein the document database stores a plurality of the vectors and indexes between the plurality of the vectors and a plurality of the document data.

6. The method of claim 1, wherein, The generating of the target question data based on the initial question data and the relevant document data comprises: combining the initial question data and the relevant document data into target question data based on a preset question template.

7. The method of claim 1, wherein, The target result data comprises first result data corresponding to the initial question data and second result data corresponding to the target question data.

8. An intelligent question answering apparatus, characterized by comprising: The method comprises: an initial question receiving module configured to receive initial question data input by a user, and determine relevant document data associated with the initial question data based on the initial question data and a pre-established document database; a target question generating module configured to generate target question data based on the initial question data and the relevant document data; an intelligent question-answering module configured to input the target question data into an intelligent question-answering model, and determine target result data based on a model output result.

9. An electronic device, comprising: The method comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the intelligent question and answer method in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to implement the intelligent question and answer method in any one of claims 1-7 when executed.

Images