Railway unfavorable geological information auxiliary question and answer implementation method and system based on large model

Abstract

The invention discloses a railway unfavorable geological information auxiliary question and answer implementation method and system based on a large model, and the method comprises the steps: firstly extracting geological industry knowledge corpora from multiple channels, constructing a pre-training industry knowledge corpus, then carrying out the incremental pre-training of a general large language model through employing the pre-training industry knowledge corpus, and carrying out the training of the large language model; and finally, in combination with business process data of the geological industry, constructing a fine-tuning industry knowledge corpus, and performing fine-tuning training on the incremental pre-training large model to form a more accurate industry knowledge large model. The constructed geological industry knowledge big model can answer basic knowledge problems of geology, provide industry consultation and assist in decision making of geological exploration and resource development, not only can perform efficient text analysis and basic knowledge questions and answers, but also can provide logic problem answers and aid decision making, and is high in practicability. And the data analysis capability and decision-making efficiency of the geological industry are obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of railway engineering technology, and in particular to a method and system for implementing auxiliary question-answering of railway adverse geological information based on a large model. Background Art

[0002] In the field of railway engineering, adverse geological conditions are one of the key factors affecting the safety, efficiency and cost of railway construction. Railway projects often face complex geological conditions, such as landslides, mudslides, karst, faults, etc. These adverse geological conditions not only increase the difficulty and cost of the project, but may also pose a serious threat to the safety of railway operations. In the traditional railway construction process, the identification and treatment of adverse geological conditions often rely on the experience and intuition of engineers. This method is not only inefficient, lacks systematic data analysis and prediction, but also may have large errors and risks. Railway engineering involves a large amount of geological data, historical records and related documents. It is a challenge to effectively integrate this information and use it for decision-making. Existing question-answering systems determine relevant answers from a large set of texts based on user questions. These replies are generally obtained by matching the user's questions. The corresponding answers are output based on the degree of matching with the user's question text, which lacks targeted answers.

[0003] In order to overcome these challenges, the railway engineering field needs to adopt more advanced technical means. First, the collection and analysis of geological data should comprehensively apply the latest artificial intelligence technologies, which can provide more detailed geological information and help engineers better understand geological conditions in all aspects. Secondly, the application of big data analysis and machine learning algorithms can identify patterns of historical geological events, predict potential geological risks, and provide data support for engineering design and construction. At the same time, the development of intelligent decision support systems, combined with expert systems and artificial intelligence technologies, can provide more accurate and real-time decision-making suggestions, thereby improving the safety and efficiency of railway engineering.

[0004] However, the application of large models in specific fields is still in its infancy and requires further training and optimization to improve its applicability and accuracy in actual engineering projects. The design of the question-answering system should take into account the actual needs of users and provide customized services to meet the specific requirements of different engineering scenarios. How to effectively integrate these poor geological data and document information to build a large data model that can comprehensively reflect poor geology is an important challenge facing the current field of poor geological data analysis. Therefore, in order to enable the large model to play a role in the real poor geological field industry, this application constructs an industry knowledge large model based on poor geological text information, thereby realizing functions such as intelligent consulting assistant for poor geological information and poor geological information analysis, which is of great significance to improving the operating efficiency and safety of geological information analysis in the railway system. Summary of the invention

[0005] The present application provides a method and system for implementing railway adverse geological information assisted question and answer based on a large model, so as to solve the problem that the existing adverse geological information question and answer system lacks targeted answers and users are easily dissatisfied with the answers output by the question and answer system.

[0006] According to the first aspect, an embodiment provides a method for implementing railway adverse geological information assisted question answering based on a large model, the method comprising:

[0007] Extract unsupervised railway geological industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geological information;

[0008] According to the different business classifications of the railway industry, the extracted data is stored in the form of classification to build a fine-tuned industry knowledge corpus;

[0009] Use pre-trained industry knowledge corpus to incrementally pre-train the general large language model;

[0010] Use the fine-tuning industry knowledge corpus to customize and fine-tune the incremental pre-trained large language model to obtain a proprietary industry knowledge large model for railway adverse geological information;

[0011] The industry knowledge big model obtained by calling in the railway adverse geological information auxiliary question and answer system provides users with industry knowledge services of adverse geological information.

[0012] Furthermore, unsupervised railway geological industry knowledge corpus is extracted from multiple channels to construct a pre-trained industry knowledge corpus containing railway adverse geological information, including:

[0013] First, determine the type and source of required data based on demand analysis;

[0014] Secondly, pre-process the data to improve the accuracy of data identification and extraction;

[0015] Thirdly, because different tools may perform differently when processing different types of files, it is necessary to select appropriate data recognition and extraction tools according to the format and characteristics of the data source. Data recognition and extraction tools include OCR recognition tools.

[0016] Finally, text recognition is performed through data recognition and extraction tools to convert the text content in the image or document into an editable format. After the recognition is completed, errors that may occur during the data extraction process are corrected, irrelevant information is removed, and key data is retained.

[0017] Furthermore, according to the different business classifications of the railway industry, the extracted data is stored in the form of classification to construct a fine-tuned industry knowledge corpus, including:

[0018] The extracted data are sorted according to the business process types of the railway industry, which include the definition of geological types, the causes of adverse geological information, adverse geological related calculations, and solutions to adverse geological information.

[0019] Furthermore, the pre-trained industry knowledge corpus is used to incrementally pre-train the general large language model, including:

[0020] The industry knowledge data in the pre-trained industry knowledge corpus is used to enhance the basic semantic understanding ability of the general large language model in the industry field; and the corpus data collected from multiple channels is put into the knowledge corpus for expansion and the corpus quantity reaches a preset scale to enhance the model's basic semantic understanding ability in the field of railway geology. The expansion method includes a tokenization method based on punctuation segmentation to convert the text into a form that the machine can understand and process.

[0021] Furthermore, the fine-tuning industry knowledge corpus is used to customize the incremental pre-trained large language model to obtain a proprietary industry knowledge large model for railway adverse geological information, including:

[0022] Fine-tune the incremental pre-trained large model according to the industry business process using industry business data, so that the incremental pre-trained large model can guide users to provide materials according to the industry business process and provide industry knowledge services to users.

[0023] Furthermore, the fine-tuning industry knowledge corpus is used to customize the incremental pre-trained large language model to obtain a proprietary industry knowledge large model for railway adverse geological information, including:

[0024] The large model fine-tuning training method based on the side network is adopted, which includes the following steps:

[0025] First, obtain the pre-trained parameters of the general large language model and the prepared fine-tuning training dataset;

[0026] Secondly, the obtained pre-trained parameters are compressed into a model to obtain the side network parameters, and the obtained side network parameters are synchronized to the GPU computing power of the local server;

[0027] Thirdly, the model is fine-tuned based on the fine-tuning training data set, wherein the pre-training parameters are fixed during fine-tuning training, forward propagation is performed locally based on the fine-tuning training data set, and the result of the forward propagation is synchronized to the external computing power; the local server computing power performs back propagation based on the result of the forward propagation and the side network parameters, and updates the side network parameters, and the local server computing power synchronizes the updated side network parameters to the local; the local server computing power synchronizes the updated side network parameters to the local environment, inputs the fine-tuning training data to each layer of the model, and weightedly adds the output of each layer of the model to the output of the corresponding layer on the side network. The result of the addition is used as the input of the side network, and the result of the forward propagation is output after calculating all the side network parameters;

[0028] Finally, after the training is completed, the final updated side network parameters are obtained locally.

[0029] Furthermore, the industry knowledge big model obtained by calling in the railway adverse geological information auxiliary question-answering system provides users with industry knowledge services of adverse geological information, including:

[0030] First, the user enters the consultation content through the question-and-answer method, and the corresponding prompt is generated;

[0031] Then, specific consultation details are asked based on the consultation content. Through the consultation content and keywords entered by the user, the trained large model is called to analyze and output relevant content for reference.

[0032] According to the second aspect, an embodiment provides a large model-based railway adverse geological information assisted question-answering implementation system, the system comprising:

[0033] A pre-trained database construction module is used to extract unsupervised railway geological industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geological information;

[0034] A fine-tuning database construction module is used to construct a fine-tuning industry knowledge corpus by storing the extracted data in a classified manner according to different business classifications of the railway industry;

[0035] The pre-training module is used to incrementally pre-train the general large language model using the pre-trained industry knowledge corpus;

[0036] The fine-tuning module is used to customize and fine-tune the incremental pre-trained large language model using the fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information;

[0037] The question-and-answer module is used to obtain the industry knowledge model through calling in the railway adverse geological information auxiliary question-and-answer system, so as to provide users with industry knowledge services of adverse geological information.

[0038] According to the third aspect, an embodiment provides an electronic device, the device comprising: a processor and a memory;

[0039] The memory is used to store one or more program instructions;

[0040] The processor is used to run one or more program instructions to execute the steps of a method for implementing a large-model-based railway adverse geological information assisted question and answer method as described in any of the above items.

[0041] According to the fourth aspect, an embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for implementing a large-model-based auxiliary question-and-answer method for railway adverse geological information are implemented as described in any of the above items.

[0042] The present application provides a method and system for implementing railway poor geological information assisted question and answer based on a large model. First, the geological industry knowledge corpus is automatically extracted from geological manuals, professional forums, and other channels, and data cleaning and formatting are performed to build a pre-trained industry knowledge corpus. Secondly, the pre-trained industry knowledge corpus is used to incrementally pre-train the general large language model to adapt it to the specific language and terminology of the geological industry. Finally, in combination with the business process data of the geological industry, a fine-tuning industry knowledge corpus is constructed, and the incremental pre-trained large model is fine-tuned to form a more accurate industry knowledge large model. The constructed geological industry knowledge large model can answer basic knowledge questions of geology, provide industry consultation, and assist in decision-making for geological exploration and resource development. The model of the present invention can not only perform efficient text analysis and basic knowledge questions and answers, but also provide logical question answers and auxiliary decision-making, significantly improving the data analysis ability and decision-making efficiency of the geological industry. The implementation of the present invention is expected to provide strong technical support for the intelligent transformation of the geological industry. The method for constructing a large model of geological industry knowledge of the present invention can effectively improve the data analysis ability and decision-making efficiency of the geological industry by combining pre-training and fine-tuning technology, and has broad application prospects and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a method for implementing a large-model-based railway adverse geological information assisted question-answering method according to an embodiment of the present invention;

[0044] Figure 2 A specific implementation flow chart of a method for implementing a large-model-based railway adverse geological information assisted question-answering method according to an embodiment of the present invention;

[0045] Figure 3A schematic diagram of the logical structure of a large-model-based railway adverse geological information assisted question-answering implementation system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0047] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0048] The first embodiment of the present invention provides a method for implementing railway adverse geological information assisted question answering based on a large model. The present invention focuses on building a large industry knowledge model (GeolGPT) in the geological field, aiming to extract valuable information from geological data and text samples, and form a geological large model that can be used online through preprocessing, pretraining and fine-tuning. Figure 1 and Figure 2 Provide detailed explanation.

[0049] like Figure 1 As shown, in step S100, unsupervised railway geology industry knowledge corpus is extracted from multiple channels to construct a pre-trained industry knowledge corpus containing railway adverse geology information.

[0050] Specifically, data extraction tools, including image recognition and text extraction tools, are used to extract unsupervised geological industry knowledge corpus from geological reports, railway engineering documents, academic papers, professional forums and other channels. A pre-trained industry knowledge corpus containing railway adverse geological information is constructed. Specifically, the sources of proprietary industry knowledge corpus for unsupervised railway adverse geology include: industry manuals, forums and other data types with railway geological conditions. The specific data can be customized according to the industry characteristics of railway geology. Data extraction tools specifically include: optical character recognition tools (OCR) and so on.

[0051] The implementation method can be implemented in the following three steps: First, according to the demand analysis, determine the type and source of the required data, such as geological reports, engineering documents, academic papers, etc. Secondly, data preprocessing to improve the accuracy of OCR recognition, including adjusting image resolution, contrast, etc. to improve recognition quality. Thirdly, since different tools may perform differently when processing different types of files (such as scanned documents, PDFs, pictures, etc.), it is necessary to select a suitable OCR tool based on the format and characteristics of the data source. The recognition accuracy of the tool, output format, etc. should be considered when selecting. This requires considering the format of the data source and the recognition ability of the OCR tool to ensure high accuracy of the conversion process. Finally, perform text recognition through the OCR tool to convert the text content in the image or document into an editable format. After the recognition is completed, correct the errors that may occur in the OCR process, remove irrelevant information, and retain key data.

[0052] like Figure 1 As shown, in step S200, according to different business classifications of the railway industry, the extracted data is stored in a classified manner to construct a fine-tuned industry knowledge corpus.

[0053] Specifically, according to the different business classifications of the railway industry, such as geological exploration, construction design, etc., the relevant data are stored in a classified manner to build a fine-tuned industry knowledge corpus.

[0054] Industry business process data includes the definition of geological types, causes of adverse geological information, adverse geological related calculations, solutions to adverse geological information, etc. This embodiment has customized settings for unique industries based on the characteristics and specific needs of the adverse geological industry, and completes the data expansion of step S100. At the same time, in the process of constructing a fine-tuning industry knowledge corpus, it is sorted according to the business process type.

[0055] like Figure 1 As shown, in step S300, the general large language model is incrementally pre-trained using a pre-trained industry knowledge corpus.

[0056] Use pre-trained industry knowledge corpus to incrementally pre-train general large language models (such as ChatGLM3-6B, LLaMA and other open source large language models). At the same time, in order to enhance the basic semantic understanding ability of the model in the field of railway geology, the number of collected corpora needs to be expanded into the knowledge corpus. The expansion result needs to reach a certain scale, such as more than 1 million tokens. Use the common punctuation-based tokenization method to convert the text into a form that the machine can understand and process.

[0057] Use the pre-trained industry knowledge corpus to incrementally pre-train the general large language model. Specifically, it can be divided into: using the industry knowledge data in the pre-trained industry knowledge corpus to enhance the basic semantic understanding ability of the general large language model in the industry field. In this step, the number of corpora must reach a certain scale to cover all business scopes as much as possible, and the scale needs to be more than 1 million tokens. In this step, the token implementation method adopts the common punctuation-based tokenization method to convert the text into a form that the machine can understand and process. In this way, the industry knowledge corpus can be effectively used to enhance the semantic understanding ability of the general large language model, making its application in specific industry fields more accurate and efficient.

[0058] like Figure 1 As shown, in step S400, the incremental pre-trained large language model is customized and fine-tuned using a fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information.

[0059] Specifically, this embodiment uses the fine-tuning industry knowledge corpus to customize the incremental pre-trained model. And build a large industry knowledge model specifically for railway adverse geological information, so that it can guide users to provide materials and provide users with accurate industry knowledge services. Fine-tuning training generally uses task-specific data sets to enhance the performance of the model on specific tasks. The above steps complete the offline training process of the large model and obtain a large model for a specific industry.

[0060] The specific content of using the fine-tuning industry knowledge corpus to customize the incremental pre-trained large model is: fine-tune the incremental pre-trained large model according to the industry business process with industry business data, so that the incremental pre-trained large model can guide users to provide materials according to the industry business process and provide industry knowledge services to users. Specifically, Prompt Engineering technology is used to guide the pre-trained language model to perform specific tasks or generate specific types of text by designing or selecting task-related prompts. The above steps complete the offline training process of the large model and obtain a large model for a specific industry.

[0061] In this embodiment, a large model fine-tuning training method based on a side network is adopted. The method is mainly divided into four steps, namely:

[0062] First, obtain the pre-trained parameters of the adopted general large language model and the prepared fine-tuning training dataset;

[0063] Secondly, the obtained pre-training parameters are compressed into the model to obtain the side network parameters. The knowledge distillation method can be used to compress the model, and the obtained side network parameters are synchronized to the GPU computing power of the local server;

[0064] Thirdly, the model is fine-tuned based on the fine-tuning training data set, wherein the pre-training parameters are fixed during fine-tuning training, forward propagation is performed locally based on the fine-tuning training data set, and the result of the forward propagation is synchronized to the external computing power; the local server computing power performs back propagation based on the result of the forward propagation and the side network parameters, and updates the side network parameters, and the local server computing power synchronizes the updated side network parameters to the local; the local server computing power synchronizes the updated side network parameters to the local environment, inputs the fine-tuning training data to each layer of the model, and weightedly adds the output of each layer of the model to the output of the corresponding layer on the side network. The result of the addition is used as the input of the side network, and the result of the forward propagation is output after calculating all the side network parameters;

[0065] Finally, after the training is completed, the final updated side network parameters are obtained locally.

[0066] like Figure 1 As shown, in step S500, the industry knowledge big model obtained by calling in the railway adverse geological information auxiliary question and answer system provides the user with industry knowledge services of adverse geological information.

[0067] Specifically, this part is model loading and service provision: loading the railway adverse geological information auxiliary question and answer system, and selecting the corresponding industry knowledge service according to different usage requirements. This part loads the large model based on the perspective of online use and then performs relevant application levels on the model.

[0068] The industry knowledge services of adverse geological information involved in this embodiment include industry consultation, text analysis, and decision support, etc. The specific implementation process is: first, the user enters the consultation content in a question-and-answer manner to generate a corresponding prompt, including: the name of the adverse geology, related calculation methods, industry-related logic consultation, etc.; then ask for specific consultation details based on the consultation content, and call the trained large model through the consultation content and keywords entered by the user to obtain analysis and output relevant content for reference. For example, for the decision-making support service, the first step is to enter and parse the Prompt material, call the trained large model, enter and parse the Prompt content, and extract material characteristics and keyword information; then there is a deep analysis of the demands, identify the demand type based on the material characteristics and keyword information, and screen the relevant knowledge information for analysis; finally, there is decision suggestion feedback, and feedback is given based on the analysis results of the industry knowledge information.

[0069] The embodiment of the present invention provides a method for implementing assisted question-answering of poor railway geological information based on a large model, which introduces the semantic analysis, text generation, and intelligent question-answering capabilities of the large model technology that are outstanding in effect into the analysis business of poor railway geological information, and constructs a large industry knowledge model for analyzing poor railway geological information by using industry knowledge corpus to incrementally pre-train and fine-tune the general large language model. At the same time, the industry knowledge large model is used to replace manual provision of geological consultation, business guidance, and decision-making assistance services, effectively reducing the workload of staff and improving decision-making efficiency. Through this method, an assisted question-answering system specifically for poor railway geological information can be constructed, which can provide accurate geological risk assessment, construction plan recommendations, and related decision-making support, thereby improving the safety and efficiency of railway projects.

[0070] Corresponding to the above-disclosed method for implementing the large-model-based railway adverse geological information assisted question-answering, the embodiment of the present invention further discloses a large-model-based railway adverse geological information assisted question-answering system, such as Figure 3 As shown, it specifically includes:

[0071] A pre-trained database construction module is used to extract unsupervised railway geological industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geological information;

[0072] A fine-tuning database construction module is used to construct a fine-tuning industry knowledge corpus by storing the extracted data in a classified manner according to different business classifications of the railway industry;

[0073] The pre-training module is used to incrementally pre-train the general large language model using the pre-trained industry knowledge corpus;

[0074] The fine-tuning module is used to customize and fine-tune the incremental pre-trained large language model using the fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information;

[0075] The question-and-answer module is used to obtain the industry knowledge model through calling in the railway adverse geological information auxiliary question-and-answer system, so as to provide users with industry knowledge services of adverse geological information.

[0076] It should be noted that for the detailed description of a system for assisted question and answer implementation of railway adverse geological information based on a large model provided in an embodiment of the present invention, reference can be made to the relevant description of a method for assisted question and answer implementation of railway adverse geological information based on a large model provided in an embodiment of the present application, which will not be repeated here.

[0077] In addition, an embodiment of the present invention also provides an electronic device, comprising: a processor and a memory; the memory is used to store one or more program instructions; the processor is used to run one or more program instructions to execute the steps of a method for implementing a large-model-based railway adverse geological information assisted question and answer method as described in any of the above items.

[0078] It should be noted that for the detailed description of an electronic device provided in an embodiment of the present invention, reference can be made to the relevant description of a method for implementing a large-model-based railway adverse geological information assisted question-and-answer method provided in an embodiment of the present application, which will not be repeated here.

[0079] In addition, an embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for implementing a large-model-based auxiliary question-and-answer method for railway adverse geological information as described in any of the above items are implemented.

[0080] It should be noted that, for the detailed description of a computer-readable storage medium provided in an embodiment of the present invention, reference can be made to the relevant description of a method for implementing a large-model-based railway adverse geological information assisted question-answering method provided in an embodiment of the present application, which will not be repeated here.

[0081] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0082] The above specific examples are used to illustrate the present invention, which are only used to help understand the present invention and are not used to limit the present invention. For those skilled in the art to which the present invention belongs, some simple deductions, modifications or substitutions can be made according to the idea of ​​the present invention.

Claims

1. A method for implementing railway adverse geological information assisted question answering based on a large model, characterized in that: The method comprises: Extract unsupervised railway geological industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geological information; According to the different business classifications of the railway industry, the extracted data is stored in the form of classification to build a fine-tuned industry knowledge corpus; Use pre-trained industry knowledge corpus to incrementally pre-train the general large language model; Use the fine-tuning industry knowledge corpus to customize and fine-tune the incremental pre-trained large language model to obtain a proprietary industry knowledge large model for railway adverse geological information; The industry knowledge big model obtained by calling in the railway adverse geological information auxiliary question and answer system provides users with industry knowledge services of adverse geological information.

2. A method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 1, characterized in that: Extract unsupervised railway geology industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geology information, including: First, determine the type and source of required data based on demand analysis; Secondly, pre-process the data to improve the accuracy of data identification and extraction; Thirdly, because different tools may perform differently when processing different types of files, it is necessary to select appropriate data recognition and extraction tools according to the format and characteristics of the data source. Data recognition and extraction tools include OCR recognition tools. Finally, text recognition is performed through data recognition and extraction tools to convert the text content in the image or document into an editable format. After the recognition is completed, errors that may occur during the data extraction process are corrected, irrelevant information is removed, and key data is retained.

3. The method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 1, characterized in that: According to the different business classifications of the railway industry, the extracted data is stored in the form of classification to build a fine-tuned industry knowledge corpus, including: The extracted data are sorted according to the business process types of the railway industry, which include the definition of geological types, the causes of adverse geological information, adverse geological related calculations, and solutions to adverse geological information.

4. The method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 1, characterized in that: Use the pre-trained industry knowledge corpus to incrementally pre-train the general large language model, including: The industry knowledge data in the pre-trained industry knowledge corpus is used to enhance the basic semantic understanding ability of the general large language model in the industry field; and the corpus data collected from multiple channels is put into the knowledge corpus for expansion and the corpus quantity reaches a preset scale to enhance the model's basic semantic understanding ability in the field of railway geology. The expansion method includes a tokenization method based on punctuation segmentation to convert the text into a form that the machine can understand and process.

5. The method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 3, characterized in that: The incremental pre-trained large language model is customized and fine-tuned using the fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information, including: Fine-tune the incremental pre-trained large model according to the industry business process using industry business data, so that the incremental pre-trained large model can guide users to provide materials according to the industry business process and provide industry knowledge services to users.

6. The method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 1, characterized in that: The incremental pre-trained large language model is customized and fine-tuned using the fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information, including: The large model fine-tuning training method based on the side network is adopted, which includes the following steps: First, obtain the pre-trained parameters of the general large language model and the prepared fine-tuning training dataset; Secondly, the obtained pre-trained parameters are compressed into a model to obtain the side network parameters, and the obtained side network parameters are synchronized to the GPU computing power of the local server; Next, the model is fine-tuned based on the fine-tuning training data set. During fine-tuning, the pre-training parameters are fixed, forward propagation is performed locally based on the fine-tuning training data set, and the results of the forward propagation are synchronized to the external computing power. The local server computing power performs back propagation based on the results of the forward propagation and the side network parameters, updates the side network parameters and synchronizes them to the local server. Finally, after the training is completed, the final updated side network parameters are obtained locally.

7. The method for implementing railway adverse geological information assisted question answering based on a large model as claimed in claim 1, characterized in that: The industry knowledge big model obtained by calling in the railway adverse geological information auxiliary question-answering system provides users with industry knowledge services of adverse geological information, including: First, the user enters the consultation content through the question-and-answer method, and the corresponding prompt is generated; Then, specific consultation details are asked based on the consultation content. Through the consultation content and keywords entered by the user, the trained large model is called to analyze and output relevant content for reference.

8. A large-scale model-based railway adverse geological information assisted question-answering system, characterized in that: The system comprises: A pre-trained database construction module is used to extract unsupervised railway geological industry knowledge corpus from multiple channels and construct a pre-trained industry knowledge corpus containing railway adverse geological information; A fine-tuning database construction module is used to construct a fine-tuning industry knowledge corpus by storing the extracted data in a classified manner according to different business classifications of the railway industry; The pre-training module is used to incrementally pre-train the general large language model using the pre-trained industry knowledge corpus; The fine-tuning module is used to customize and fine-tune the incremental pre-trained large language model using the fine-tuning industry knowledge corpus to obtain a proprietary industry knowledge large model for railway adverse geological information; The question-and-answer module is used to obtain the industry knowledge model through calling in the railway adverse geological information auxiliary question-and-answer system, so as to provide users with industry knowledge services of adverse geological information.

9. An electronic device, characterized in that: The device comprises: a processor and a memory; The memory is used to store one or more program instructions; The processor is used to run one or more program instructions to execute the steps of a method for implementing a large-model-based railway adverse geological information assisted question and answer method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for implementing a large-model-based railway adverse geological information assisted question-answering method as described in any one of claims 1 to 7.

Images