Intelligent question answering system and question answering method for professional knowledge of tobacco industry based on big language

Abstract

The invention relates to an intelligent question-answering system and question-answering method for tobacco industry professional knowledge based on a big language, and the system comprises a training data collection unit which is used for collecting related professional knowledge and related work flow and system files of the tobacco industry, and carrying out the deep learning and training, and forming a knowledge base; the model construction unit is used for constructing a large language model platform; the user interaction unit is used for performing information interaction between employees and the constructed large language model platform; the text question and answer unit is provided with an answer text acquisition module, a question generation module and a preset dialogue generation model; the knowledge base access module is used for accessing the employee question and answer text as a knowledge base to the large language model platform through a preset dialogue generation model; and the matching answer output module is used for searching answers matched with the questions from the knowledge base and outputting the matched answers. Through combination with an existing business process, a large language model privatization deployment platform is constructed, internal enterprise WeChat is connected, and intelligent question answering and pushing are formed.

Description

Technical Field

[0001] The present application relates to the relevant technical fields of natural language processing, and in particular to an intelligent question-answering system and a question-answering method based on big language and tobacco industry professional knowledge. Background Art

[0002] In related technologies, question-answering systems are one of the applications of natural language processing technology in real-world scenarios, and are widely used in various fields such as law, medicine, and finance. Common vertical question-answering systems require the use of a specified document collection to answer user questions, so they are also called knowledge-based question-answering systems.

[0003] With the continuous deepening of digitalization, informatization and intelligence in the tobacco industry, the application fields of the tobacco industry are constantly expanding, and the resulting tobacco data formats are complex and diverse. In addition to conventional structured data, there are also a large number of semi-structured and unstructured data; these data, as professional knowledge of the tobacco industry, together constitute a huge, scattered, multi-source, heterogeneous, multi-dimensional and multi-form tobacco data resource in a multimodal form. The postures contained in these resources can be utilized by intelligent question-answering systems.

[0004] However, through research in the tobacco industry, the existing technical solutions are relatively simple. Usually, the relevant learning materials are placed on a network disk. After new employees or transferred employees arrive at a new department, they log in to the network disk to search for relevant policy or system documents based on the file name in the directory. They can only search through their own cognition and memory, or hope that other colleagues can provide help, which takes a long time and is not necessarily accurate. The actual operation efficiency is low. The network disk technology is just a directory for shared files, and files can only be found by people actively, resulting in low work efficiency and high error rate of employees. Summary of the invention

[0005] In order to overcome the problems existing in the relevant technologies, this application provides an intelligent question-answering system and a question-answering method based on big language tobacco industry professional knowledge. Through the deployment and promotion of this knowledge base, it can help new employees, transferred employees, or employees who lack business familiarity to quickly and accurately answer business knowledge and operation points in the execution of specific work tasks, improve the efficiency and standardization of the execution of related work tasks, help improve the efficiency of task advancement of personnel in various positions, and reduce the probability of work errors.

[0006] The first aspect of the present application provides an intelligent question-answering system based on big language tobacco industry professional knowledge, including:

[0007] The training data collection unit is used to collect relevant professional knowledge of the tobacco industry and related work processes and system documents, and conduct deep learning and training to form a knowledge base;

[0008] Model building unit, building a large language model platform;

[0009] User interaction unit, used for employees to interact with the constructed large language model platform;

[0010] The text question and answer unit is provided with an answer text acquisition module and a question generation module, and a preset dialogue generation model; a knowledge base access module is used to access the employee question and answer text as a knowledge base to the large language model platform through the preset dialogue generation model; a matching answer output module is used to obtain an input question, search for an answer matching the question from the knowledge base through the large language model platform, and output the matching answer.

[0011] In the above-mentioned intelligent question-answering system based on big language and tobacco industry professional knowledge, the intelligent question-answering system uses a hybrid mode of the Bert mode and the GPT mode as well as a big language model and knowledge retrieval to build the underlying platform.

[0012] In the above-mentioned intelligent question-and-answer system based on big language and professional knowledge of the tobacco industry, the intelligent question-and-answer system is built on the WEB side using C# development technology based on the underlying platform, and cooperates with the human-computer interaction mode of the mobile enterprise WeChat.

[0013] In the above-mentioned intelligent question-answering system based on big language and tobacco industry professional knowledge, the construction of the underlying platform includes: installing the basic environment; building a big language model and knowledge retrieval platform; importing the models chatGLM2-6B, moka-ai / m3e-base and configuring them; importing and training knowledge base materials; installing the application environment .Net Core package, Nginx, mysql; building the application; configuring the application to call the relevant configuration of the big language model platform; and completing the construction.

[0014] In the above-mentioned intelligent question-answering system based on big language and tobacco industry professional knowledge, the intelligent question-answering system is a privately deployed system architecture, the constructed knowledge base is localized data, and the rtst mode and the fess mode are used as knowledge base search models.

[0015] In the above-mentioned intelligent question-and-answer system based on big language and professional knowledge of the tobacco industry, the system architecture includes a data layer, a language model, and an application layer; the data layer is provided with vector data, document data, cache data, and business data; the language model is provided with a knowledge base, a big language model, and basic technical capabilities; the application layer includes a web page and an enterprise WeChat applet, including knowledge questions and answers, document parsing, copywriting, user management, knowledge base maintenance, parameter settings, and system logs.

[0016] In the above-mentioned intelligent question-answering system based on big language and tobacco industry professional knowledge, in the language model, the knowledge base includes a local offline vector library, a local search engine, and an online search engine; the big language model is chatGLM-6B and its core technology; the basic technical capabilities include parameter tuning, cleaning knowledge base files, and RTST knowledge base maintenance.

[0017] In the above-mentioned intelligent question-and-answer system based on big language and professional knowledge of the tobacco industry, the user interaction unit is developed using the enterprise WeChat of the user's office. The user interacts with the program through the enterprise WeChat server. When the user interacts with the application in the enterprise WeChat, the enterprise WeChat server is first accessed. When accessing the enterprise WeChat server, the application key and callback address will be brought with it. After the verification result, a callback is made to access the application address. When the callback is made, the enterprise WeChat brings a verification code, through which the user's information is obtained, and then the user information is queried and exchanged with the database.

[0018] The second aspect of the present application provides an intelligent question-answering method based on tobacco industry professional knowledge in a big language, comprising:

[0019] S1. Collect original dialogue materials of tobacco industry professional knowledge and build a tobacco industry professional knowledge database. Through the tobacco industry professional knowledge database, align the dialogue scenarios in the process of intelligent question-answering of tobacco industry professional knowledge, form an answer text acquisition module and a question generation module, and preset a dialogue generation model.

[0020] S2. Build a large language model platform for the tobacco industry professional knowledge intelligent question-answering system; use the tobacco industry professional knowledge base to perform migration and adaptation training on the large language model platform in the tobacco industry, and evaluate the trained large language model;

[0021] S3. The tobacco industry professional knowledge base is mounted on the evaluated large language model platform to complete the construction of the intelligent question-answering system. During the intelligent question-answering process, the input question and the tobacco industry professional knowledge base are converted to the same vector space, and the matching results are retrieved and calculated. The large language model aligns the dialogue scenario of the input question, applies the trained language ability and understanding integration ability, and describes the matching result as the output answer.

[0022] A third aspect of the present application provides an electronic device, including:

[0023] Processor; and

[0024] The memory stores executable codes thereon, and when the executable codes are executed by the processor, the processor is caused to execute the method as described above.

[0025] The technical solution provided in this application may include the following beneficial effects: the intelligent question-and-answer system and question-and-answer method based on the tobacco industry professional knowledge of the big language, by combining with the existing business process, constructing a large language model privatization deployment platform, connecting to the internal enterprise WeChat, forming intelligent question-and-answer and push. At the same time, a knowledge base of documents such as relevant processes, systems and other documents of the company's internal procurement, finance, monopoly, audit, personnel and other departments is established. Compared with the data consultation and material learning of personnel office under the traditional business model, intelligent question-and-answer is realized, and the daily office efficiency of personnel is improved. By launching the big language model platform application, internal personnel can obtain the knowledge content they want to obtain more quickly, and can allow new employees or transferred employees to quickly find relevant knowledge in their new positions. The content covers various business areas, improves employee work efficiency and meets personal learning needs. At the same time, for new employees or transferred employees, it is urgent to obtain accurate business knowledge of the system in the new position to adapt to the new position faster. The knowledge base helps them to obtain business knowledge of the new position, keep abreast of the professional and related development trends involved in the new position, save time spent on information acquisition, and thus improve work efficiency.

[0026] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and other objects, features and advantages of the present application will become more apparent through a more detailed description of exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0028] Figure 1 It is a structural schematic diagram of an intelligent question-answering system based on tobacco industry professional knowledge in a big language shown in an embodiment of the present application;

[0029] Figure 2 It is a flowchart of an intelligent question-answering method based on tobacco industry professional knowledge in a big language shown in an embodiment of the present application;

[0030] Figure 3 It is a structural diagram of the system architecture of the intelligent question-answering system based on tobacco industry professional knowledge in a big language shown in an embodiment of the present application;

[0031] Figure 4 It is a flowchart of an intelligent question-answering method based on tobacco industry professional knowledge in a big language shown in an embodiment of the present application;

[0032] Figure 5 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application. DETAILED DESCRIPTION

[0033] The preferred embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0034] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0035] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0036] The technical solution of the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0037] See also Figure 1 , an intelligent question and answer system for tobacco industry professional knowledge based on big language, including: a training data acquisition unit, a model building unit, a text question and answer unit, a knowledge base access module, and a knowledge base access module.

[0038] The training data collection unit is used to collect relevant professional knowledge of the tobacco industry as well as relevant work processes and system documents, and conduct deep learning and training to form a knowledge base.

[0039] The model building unit builds a large language model platform. The user interaction unit is used for employees to interact with the built large language model platform.

[0040] The text question and answer unit is provided with an answer text acquisition module and a question generation module, and a preset dialogue generation model. The answer text acquisition module is used to receive an employee question and answer implementation request, and obtain the employee answer text according to the employee question and answer implementation request, and the employee answer text includes multiple paragraphs of content. The question generation module is used to generate one or more corresponding questions for each paragraph of content according to the employee answer text through a large language model to obtain an employee question and answer text, and the employee question and answer text is processed into a question and answer format. The knowledge base access module is used to access the employee question and answer text as a knowledge base to the large language model platform through a preset dialogue generation model. The matching answer output module is used to obtain an input question, search for an answer matching the question from the knowledge base through the large language model platform, and output the matching answer.

[0041] This intelligent question-and-answer system for tobacco industry professional knowledge based on big language studies the LLM big language model, selects the LLM big language model based on the adaptability requirements of industry positions, and builds an artificial intelligence plug-in knowledge base involving various work systems and process materials within the tobacco industry. In the form of questions and answers, it realizes the self-use and self-receiving of process skills and business knowledge corresponding to work projects in various positions within the industry, helping new employees or transferred employees to adapt to new positions more quickly and efficiently, standardize business processes, improve management levels, and achieve high-quality development and digital transformation of related businesses. Through the deployment and promotion of this knowledge base, it can help new employees, transferred employees, or those who lack familiarity with the business to achieve fast and accurate self-service questions and answers on business knowledge and key points in the execution of specific work tasks, improve the efficiency and standardization of the execution of related work tasks, help improve the efficiency of task advancement of personnel in various positions, and reduce the probability of work errors.

[0042] In the consideration of the job performance capabilities of existing personnel in various positions in the industry, a complete set of pre-job training processes are designed for new personnel, especially newly employed college students, to ensure that they can adapt to the job requirements as soon as possible. However, the overall training content is mostly related to the industry's main business, which is biased towards comprehensive quality training. The operating skills and business knowledge involved in the professional directions of each specific position are relatively scattered, and the training can only cover newly employed college students, not those who have changed positions. Therefore, in order to improve the industry's employment efficiency and control employment costs, it is necessary to find a faster and more efficient information transmission path that has both the business knowledge and operating skills required by the job profession.

[0043] In addition, when new or transferred employees forward or transmit relevant documents through the OA system in their daily work, they often encounter problems such as scattered information, insufficient timeliness, and nowhere to find. Once there is a need to find specific policy or system documents at a specific time, employees can only search through their own cognition and memory, or hope that other colleagues can provide help, which takes a long time and is not necessarily accurate, and the actual operation efficiency is low. By building a knowledge base covering various work systems, processes and various materials in the tobacco industry, new or transferred employees can retrieve and ask questions more conveniently, quickly and accurately, and can quickly adapt to new jobs, making business handovers and information transmission more accurate, convenient and efficient.

[0044] The technical routes adopted by the application of large language model platforms can be mainly divided into Bert mode, GPT mode, hybrid mode and large language model (LLM) + knowledge retrieval. Among them, Bert mode and GPT mode are two mainstream technical routes. Bert mode adopts a two-stage method, namely bidirectional language model pre-training and task fine-tuning, which is suitable for understanding, doing understanding, and specific tasks of a certain scene, which is specialized and light. The GPT mode has obvious advantages in data type, task type and model scale, and is regarded as a more general technical route. In addition, most of them adopt a hybrid mode, that is, combining the advantages of Bert mode and GPT mode to improve the application effect of large language models. In addition, the large language model (LLM) + knowledge retrieval method is also widely used, which can make up for the lack of basis and hallucinations in the answers of general large language models in professional fields. This intelligent question-answering system uses a hybrid mode of the Bert mode and the GPT mode, as well as a large language model and knowledge retrieval to build the underlying platform. Based on the underlying platform, C# development technology is used to build the WEB side, and the human-computer interaction mode of the mobile enterprise WeChat is coordinated.

[0045] The construction of the underlying platform includes: installing the basic environment; building a large language model and knowledge retrieval platform; importing the models chatGLM2-6B, moka-ai / m3e-base and configuring them; importing and training knowledge base materials; installing the application environment .Net Core package, Nginx, and mysql; building an application; configuring the application to call the relevant configuration of the large language model platform; and completing the construction.

[0046] Large language model platform, supporting knowledge base including local offline vector library, local search engine, online search engine, etc. Integrate multiple large language modules including chatGLM-6B\chatGLM2-6B, chatRWKV, Aquila-7B, InternLM, online API access openai api and chatGLM-130b api, etc. Support Auto script, by developing plug-in JavaScript scripts, add functions to the platform, including but not limited to custom dialogue process, access to external API, online switching of LoRA model.

[0047] The knowledge base of the large language model is constructed. The principle of the knowledge base is to generate some prompt information and insert it into the dialogue after the search, so that the data of the knowledge base is known by the model. The rtst mode calculates semantics and matches in the local database; the fess mode (equivalent to the local search engine) and the bing mode both call the search engine to search for answers. The construction adopts the following modes: rtst mode, sentence_transformers+faiss for indexing, supports pre-built indexes and running builds. Bing mode, cn.bing search. Bingsite mode, cn.bing site search. Fess mode, locally deployed fess search, and keyword extraction.

[0048] Search performance is improved by optimizing the lora weight fine-tuning model and combining the large language model fess search module. Fess is a powerful and easy-to-deploy enterprise search server. Fess can be quickly installed and run on any platform that can run the Java Runtime Environment. Fess is based on the OpenSearch / Elasticsearch search engine. At the same time, Fess provides an easy-to-use management GUI that is easy to configure on demand. Fess also includes quick search for documents on the file system or data storage (such as CSV or database). Supports multiple file formats, including (but not limited to): Microsoft Office, PDF, and zip, etc. The intelligent question and answer system is a privately deployed system architecture. The knowledge base constructed is localized data, and the rtst mode and fess mode are used as knowledge base search models.

[0049] See also Figure 3The system architecture includes a data layer, a language model, and an application layer; the data layer is provided with vector data, document data, cache data, and business data; the language model is provided with a knowledge base, a large language model, and basic technical capabilities; the application layer includes a web page and an enterprise WeChat applet, including knowledge questions and answers, document parsing, copywriting, user management, knowledge base maintenance, parameter settings, and system logs.

[0050] In the language model, the knowledge base includes a local offline vector library, a local search engine, and an online search engine; the large language model is chatGLM-6B and its core technology; the basic technical capabilities include parameter tuning, cleaning knowledge base files, and RTST knowledge base maintenance.

[0051] See also Figure 4 The user interaction unit is developed using the enterprise WeChat used by users for office work. The user interacts with the program through the enterprise WeChat server. When a user interacts with an application in enterprise WeChat, the enterprise WeChat server is accessed first. When accessing the enterprise WeChat server, the application key and callback address are carried with the server. After verification, a callback is made to access the application address. When the callback is made, enterprise WeChat carries a verification code, through which the user's information is obtained, and then the user information is queried and exchanged with the database.

[0052] See also Figure 2 , an intelligent question-answering method based on big language tobacco industry expertise, including:

[0053] S1. Collect original dialogue materials of tobacco industry professional knowledge and build a tobacco industry professional knowledge database. Through the tobacco industry professional knowledge database, align the dialogue scenarios in the process of intelligent question-answering of tobacco industry professional knowledge, form an answer text acquisition module and a question generation module, and preset a dialogue generation model.

[0054] S2. Build a large language model platform for the tobacco industry professional knowledge intelligent question-answering system; use the tobacco industry professional knowledge base to perform migration and adaptation training on the large language model platform in the tobacco industry, and evaluate the trained large language model;

[0055] S3. The tobacco industry professional knowledge base is mounted on the evaluated large language model platform to complete the construction of the intelligent question-answering system. During the intelligent question-answering process, the input question and the tobacco industry professional knowledge base are converted to the same vector space, and the matching results are retrieved and calculated. The large language model aligns the dialogue scenario of the input question, applies the trained language ability and understanding integration ability, and describes the matching result as the output answer.

[0056] The traditional office mode is upgraded by using open source big language model technology, and artificial intelligence and deep learning technology are used to integrate the system and process of the company's procurement, finance, monopoly, audit, personnel and other work into the internal knowledge base, and finally form a reliable, accurate and efficient office intelligent question and answer system. By combining with the existing business process, a large language model private deployment platform is built, which is connected to the internal enterprise WeChat to form intelligent question and answer and push. Compared with the traditional business model of data consultation and material learning of personnel office, intelligent question and answer is realized, which improves the daily office efficiency of personnel and is expected to save 2 hours / person-day. Through this intelligent question and answer system based on big language professional knowledge of the tobacco industry, internal personnel can obtain the knowledge content they want more quickly, and new employees or transferred employees can quickly find relevant knowledge in new positions. The content covers various business areas, improves employee work efficiency and meets personal learning needs. At the same time, for new employees or transferred employees, it is urgent to obtain accurate business knowledge in the new position to adapt to the new position faster. The knowledge base helps them to obtain business knowledge of the new position, keep abreast of the professional and related development trends involved in the new position, save time spent on information acquisition, and thus improve work efficiency.

[0057] See also Figure 5 , the electronic device 1000 includes a memory 1010 and a processor 1020 .

[0058] The processor 1020 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0059] The memory 1010 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. Among them, ROM can store static data or instructions required by the processor 1020 or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at run time. In addition, the memory 1010 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory 1010 may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (such as a DVD-ROM, a double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a mini SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0060] The memory 1010 stores executable codes, and when the executable codes are processed by the processor 1020 , the processor 1020 can execute part or all of the methods described above.

[0061] The scheme of the present application has been described in detail above with reference to the accompanying drawings. In the above embodiments, the description of each embodiment has its own emphasis. For the part that is not described in detail in a certain embodiment, refer to the relevant description of other embodiments. Those skilled in the art should also know that the actions and modules involved in the description are not necessarily required for the present application. In addition, it can be understood that the steps in the method of the embodiment of the present application can be adjusted in order, merged and deleted according to actual needs, and the modules in the device of the embodiment of the present application can be merged, divided and deleted according to actual needs.

[0062] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0063] Alternatively, the present application can also be implemented as a non-temporary machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) on which executable code (or computer program, or computer instruction code) is stored. When the executable code (or computer program, or computer instruction code) is executed by a processor of an electronic device (or electronic device, server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0064] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the application herein may be implemented as electronic hardware, computer software, or combinations of both.

[0065] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system and method according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, a program segment or a code, and the part of the module, the program segment or the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or the flow chart, and the combination of the square boxes in the block diagram and / or the flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0066] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent question-answering system based on big language and tobacco industry expertise, It is characterized in that include: The training data collection unit is used to collect relevant professional knowledge of the tobacco industry and related work processes and system documents, and conduct deep learning and training to form a knowledge base; Model building unit, building a large language model platform; User interaction unit, used for employees to interact with the constructed large language model platform; The text question-answering unit has an answer text acquisition module and a question generation module, and a preset dialogue generation model; The knowledge base access module is used to access the employee question and answer text as a knowledge base to the large language model platform through a preset dialogue generation model; the matching answer output module is used to obtain an input question, search for an answer matching the question from the knowledge base through the large language model platform, and output the matching answer.

2. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 1, Features: The intelligent question-answering system uses a hybrid mode of the Bert mode and the GPT mode as well as a large language model and knowledge retrieval to build the underlying platform.

3. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 2, Features: The intelligent question-and-answer system is built on the WEB side using C# development technology based on the underlying platform, and is coordinated with the human-computer interaction mode of the mobile enterprise WeChat.

4. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 2, Features: The construction of the underlying platform includes: installing the basic environment; building a large language model and knowledge retrieval platform; importing the models chatGLM2-6B, moka-ai / m3e-base and configuring them; importing and training knowledge base materials; installing the application environment .Net Core package, Nginx, and mysql; building an application; configuring the application to call the relevant configuration of the large language model platform; and completing the construction.

5. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 1, Features: The intelligent question-answering system is a privately deployed system architecture. The constructed knowledge base is localized data, and the rtst mode and the fess mode are used as knowledge base search models.

6. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 5, Features: The system architecture includes a data layer, a language model, and an application layer; the data layer is provided with vector data, document data, cache data, and business data; the language model is provided with a knowledge base, a large language model, and basic technical capabilities; the application layer includes a web page and an enterprise WeChat applet, including knowledge questions and answers, document parsing, copywriting, user management, knowledge base maintenance, parameter settings, and system logs.

7. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 6, Features: In the language model, the knowledge base includes a local offline vector library, a local search engine, and an online search engine; the large language model is chatGLM-6B and its core technology; the basic technical capabilities include parameter tuning, cleaning knowledge base files, and RTST knowledge base maintenance.

8. The intelligent question-answering system based on tobacco industry expertise in big language according to claim 1, Features: The user interaction unit is developed using WeChat for Enterprise, where users interact with programs. When users interact with applications in WeChat for Enterprise, they first access the WeChat for Enterprise server. When accessing the WeChat for Enterprise server, they carry the application key and callback address. After verification, a callback is made to access the application address. When the callback is made, WeChat for Enterprise carries a verification code, through which the user's information is obtained, and then data is queried and exchanged through the user information and the database.

9. An intelligent question-answering method based on big language and tobacco industry expertise, It is characterized in that include: S1. Collect original dialogue materials of tobacco industry professional knowledge and build a tobacco industry professional knowledge database; Through the tobacco industry professional knowledge database, the conversation scenarios in the process of intelligent question-answering of tobacco industry professional knowledge are aligned to form an answer text acquisition module and a question generation module, and a preset conversation generation model; S2. Build a large language model platform for the tobacco industry professional knowledge intelligent question-answering system; use the tobacco industry professional knowledge base to perform migration and adaptation training on the large language model platform in the tobacco industry, and evaluate the trained large language model; S3. The tobacco industry professional knowledge base is mounted on the large language model platform after evaluation to complete the construction of the intelligent question-answering system; In the intelligent question-answering process, the input question and the tobacco industry professional knowledge base are converted into the same vector space, and the matching results are retrieved and calculated; The large language model aligns the dialogue scenario of the input question, applies the trained language ability and understanding integration ability, and describes the matching result as the output answer.

10. An electronic device, It is characterized in that include: processor; as well as A memory having executable codes stored thereon, which, when executed by the processor, causes the processor to execute the system or method according to any one of claims 1 to 9.

Images