Fault short message semantic analysis method and device based on multi-language dialogue model

Abstract

The invention provides a fault short message semantic analysis method and device based on a multi-language dialogue model. The method comprises the following steps: a fault short message knowledge data storage step; a fault short message knowledge data preprocessing step: preprocessing the fault short message knowledge data, and segmenting long text data into smaller processable units; a data vector conversion step: calling an Embedding model by utilizing a Langchain framework, converting the preprocessed knowledge data into vector representation, storing vectorized data into a vector database, and storing and managing the vector data by adopting a Faise vector database; in the data vector retrieval step, similar historical fault cases are retrieved from the Faiss vector database, and fault related nodes and related data with the highest correlation are quickly found through vector similarity calculation and result rearrangement. According to the invention, fault short messages can be received and processed in real time, a fault source can be quickly positioned, an instant solution is provided, and the accuracy and efficiency of fault processing are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data processing, and in particular relates to a method and device for semantic parsing of fault text messages based on a multi-language dialogue model. Background Art

[0002] In recent years, with the rapid development of the civil aviation industry, civil aviation passenger service systems have played an important role in flight management, ticket processing, etc. However, as the complexity of the system continues to increase, system failures have become more frequent and difficult to predict. Traditional troubleshooting and handling methods usually rely on manual experience, which is not only time-consuming and labor-intensive, but also prone to misjudgment and delays. Therefore, how to efficiently and accurately locate the source of the fault has become a key issue to ensure the stable operation of the system.

[0003] In this context, fault messages, as the first-hand information when a fault occurs, have become an important data source for quickly locating and handling faults. However, since fault messages are expressed in a variety of ways and their content is complex and changeable, it is challenging to directly extract key information from them for fault analysis and location. In addition, the associations between the various software modules in the civil aviation passenger service system are complex, and how to accurately determine the software modules related to the fault is also a difficult problem currently faced. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a semantic parsing device for fault SMS based on the multilingual dialogue model ChatGLM3-6B and retrieval enhanced generation (RAG) technology. The device uses the historical data of fault SMS of the civil aviation passenger service system and the knowledge graph data of the software responsibilities and the association relationship between software of the civil aviation passenger service system as the knowledge base. When a system failure occurs, the device can quickly locate the relevant fault information by parsing the fault SMS and predict the software module that may fail.

[0005] The present invention provides a method for semantic parsing of fault text messages based on a multilingual dialogue model, the method comprising:

[0006] A fault SMS knowledge data storage step, storing and storing previous fault SMS knowledge data in the system;

[0007] The fault SMS knowledge data preprocessing step preprocesses the fault SMS knowledge data, including data cleaning and block operations, so as to divide the long text data into smaller processable units, thereby improving the efficiency and accuracy of subsequent processing;

[0008] In the data vector conversion step, the Embedding model is called using the Langchain architecture to convert the preprocessed knowledge data into vector representation, and the vectorized data is stored in the vector database. The Faiss vector database is used to store and manage these vector data.

[0009] In the data vector retrieval step, the Embedding model and the Rerank model are called to retrieve similar historical fault cases from the Faiss vector database. The fault-related nodes and related data with the highest correlation are quickly found through vector similarity calculation and result re-ranking.

[0010] Furthermore, the method further comprises the following steps:

[0011] The steps for uploading fault SMS knowledge include uploading the knowledge graph data of fault SMS history data, software responsibilities, and the relationship between software.

[0012] Furthermore, the method further comprises the following steps:

[0013] Fault SMS knowledge data monitoring step: The interface monitors the fault SMS content and sends the monitored content to the fault SMS knowledge uploading step.

[0014] Furthermore, the fault SMS history data includes historical fault SMS content and processing solutions, the software responsibilities mainly describe the software functions and architecture of the civil aviation passenger service system, and the knowledge graph data of the relationship between software includes the upstream and downstream relationships between various software in the civil aviation passenger service system.

[0015] Furthermore, the fault SMS knowledge data preprocessing step includes: using Python's third-party libraries python-docx, pdfplumber and pandas to divide the above long text data into smaller text blocks for subsequent processing, and the cutting parameters are 100 characters per block, with 20 characters overlapping above and below.

[0016] Furthermore, the data vector retrieval step includes: using a Rerank model to re-rank the retrieved candidate results by relevance to obtain the top five most relevant fault information, wherein the most relevant fault information includes fault type, relevant software modules and historical solutions.

[0017] Furthermore, the relevance reordering includes: assembling and generating a Prompt according to the reordering result, wherein the Prompt includes: the name of the faulty software, the fault solution and the content of the received fault SMS message.

[0018] Furthermore, the generated Prompt is utilized, the multi-language dialogue model is ChatGLM3-6B, ChatGLM3-6B is called, the fault SMS is semantically parsed and a final result is generated, the final result including the software name that may fail and the corresponding solution suggestion.

[0019] The present invention also provides a fault SMS semantic parsing device based on a multi-language dialogue model, the device comprising:

[0020] Fault SMS knowledge data storage module, which stores and manages the previous fault SMS knowledge data in the system;

[0021] The fault SMS knowledge data preprocessing module preprocesses the fault SMS knowledge data, including data cleaning and block operations, dividing the long text data into smaller processable units to improve the efficiency and accuracy of subsequent processing;

[0022] The data vector conversion module uses the Langchain architecture to call the Embedding model, converts the preprocessed knowledge data into vector representation, stores the vectorized data into the vector database, and uses the Faiss vector database to store and manage these vector data;

[0023] The data vector retrieval module calls the Embedding model and the Rerank model to retrieve similar historical fault cases from the Faiss vector database, and quickly finds the most relevant fault-related nodes and related data through vector similarity calculation and result re-ranking.

[0024] The present invention also provides a computer-readable storage medium storing one or more programs, which, when executed, can implement the aforementioned fault SMS semantic parsing method.

[0025] Compared with the prior art, the present invention has the following advantages:

[0026] The present invention has significant advantages in analyzing and processing fault SMS messages in civil aviation passenger service systems, specifically including the following aspects:

[0027] 1. Automated fault analysis: This technical solution analyzes fault SMS messages through automated means, greatly reducing the workload and time of manual analysis. The system can receive and process fault SMS messages in real time, quickly locate the source of the fault, provide immediate solutions, and improve the accuracy and efficiency of fault handling.

[0028] 2. High-precision retrieval and sorting: Through the combination of Embedding models (such as m3e-base) and Faiss vector database, high-precision vector retrieval and sorting are achieved. The Rerank model is used to re-rank the retrieval results to ensure that the most relevant historical fault information and solutions are provided, thereby improving the accuracy of fault diagnosis.

[0029] 3. Intelligent knowledge maintenance: This solution includes a fault SMS knowledge maintenance module, which can manage and maintain the knowledge graph data of fault SMS history data, software responsibilities and the relationship between software. The system pre-processes and vectorizes these knowledge data to ensure timely update and efficient use of the knowledge base.

[0030] 4. Multi-language support: The models used in this technical solution (such as ChatGLM-6B and m3e-base) have multi-language support capabilities and can handle the needs of fault SMS analysis in a multi-language environment. This is of great significance to the international civil aviation system and can adapt to the fault handling requirements in different language environments.

[0031] 5. Flexibility and scalability: The system is designed to be flexible and can be adjusted and expanded according to specific needs. For example, the system performance can be improved by replacing different embedding models or vector databases, or new functional modules can be added to meet specific needs. Such a design ensures the sustainable development and upgrade of the system.

[0032] 6. Real-time response: The system can receive and process fault messages in real time and provide instant fault diagnosis and solutions, significantly shortening the response time of fault handling. For systems with high real-time requirements such as the civil aviation passenger service system, it can effectively reduce the impact of faults and ensure the stable operation of the system.

[0033] 7. Data-driven decision support: Through systematic analysis of fault SMS and historical data, this technical solution can provide data-driven decision support. The system can not only quickly locate faults, but also provide trend analysis and early warning information by analyzing historical fault data, providing a scientific basis for system maintenance and optimization.

[0034] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A module diagram of a fault SMS semantic parsing system based on a multi-language dialogue model is shown;

[0037] Figure 2 A semantic parsing process of fault SMS based on a multilingual dialogue model is shown;

[0038] Figure 3 A front-end diagram of a fault SMS knowledge maintenance module based on a multi-language dialogue model is shown. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] The ChatGLM3-6B model is an advanced natural language processing model with powerful semantic understanding and generation capabilities. By combining RAG technology, the device can not only extract key information from fault text messages, but also provide more accurate and detailed fault information by comparing and correlating data with the knowledge base. This method not only improves the speed and accuracy of fault location, but also provides an important reference for subsequent fault handling.

[0041] The existing civil aviation passenger service system troubleshooting mainly relies on manual analysis, which is not only time-consuming and labor-intensive, but also prone to misjudgment, resulting in low efficiency in fault handling. To address this problem, the present invention proposes an intelligent semantic parsing device for fault SMS based on the ChatGLM3-6B model and retrieval enhanced generation (RAG) technology. The device can automatically parse the key information in the fault SMS, combine the knowledge graph data of the system software responsibilities and the relationship between software, quickly locate the software modules that may fail, improve the accuracy and efficiency of fault handling, and thus significantly improve the operational stability of the system.

[0042] Before describing the present invention in detail, the professional terms used in the present invention and their explanations are introduced first.

[0043] Large language models: Large language models refer to complex artificial intelligence systems that can process and generate natural language text.

[0044] RAG technology: RAG (Retrieval-Augmented Generation) technology is a technology that combines information retrieval (IR) and natural language generation (NLG). Its core idea is to introduce the content of external knowledge base as a supplement in the process of generating text to enhance the performance of the generation model.

[0045] ChatGLM-6B: A large language model with 7 billion parameters developed by the Knowledge Engineering and Data Mining Laboratory of Tsinghua University.

[0046] Langchain: LangChain is an open source framework for building applications that use language models, such as large language models, to perform various tasks. It aims to simplify the development process and allow developers to easily integrate and utilize advanced language model capabilities.

[0047] Chunk: In text processing, long documents may be split into smaller "chunks" for easier processing or analysis. This is common in natural language processing (NLP), especially when dealing with large models, such as splitting text into paragraphs that fit the model input length. Parameters: chunk_size The maximum number of characters per chunk, chunk_overlap The number of overlapping characters per chunk.

[0048] Embedding model: Embedding model is a technology used in natural language processing (NLP) and other machine learning tasks, which converts text or symbols into numerical vectors. These vectors capture the inherent structure and semantic information of the original data, enabling machine learning algorithms to understand and process language data.

[0049] Rerank model: Rerank model is a technique used in the fields of information retrieval and natural language processing to improve the quality of initial retrieval results. Such models are usually used to re-rank the list of documents retrieved initially to ensure that the most relevant documents are at the top of the result list.

[0050] DataFrame: A data structure used to store tabular data, commonly found in data analysis and data science libraries.

[0051] Faiss: Faiss (Facebook AI Similarity Search) is a library for efficient similarity search and clustering of large-scale collections, suitable for processing high-dimensional vectors. The main purpose is to accelerate similarity search of large-scale vector collections. It supports a variety of index structures and distance measurement methods, and can run on CPU and GPU.

[0052] m3e-base: Developed by the MokaAI team, the M3E model aims to unify various text embedding functions, supports multiple languages ​​including Chinese and English, and can handle input data of different granularities.

[0053] bge-reranker-v2-m3: A multi-lingual and multi-functional ranking model developed by Beijing Academy of Artificial Intelligence (BAAI).

[0054] System composition

[0055] The fault SMS semantic parsing device of the present invention is based on the civil aviation passenger service system, such as Figure 1 As shown, it mainly includes the following modules: fault SMS knowledge maintenance module, knowledge data preprocessing module, data vector module, vector database, vector retrieval module, and large language model base.

[0056] 1. Fault SMS knowledge maintenance module:

[0057] ·Fault SMS historical data: Store and manage the previous fault SMS data in the system. These data provide the system with a wealth of fault troubleshooting, location and solution cases.

[0058] Software responsibilities: Clarify the functions and responsibilities of each software module to facilitate accurate analysis and troubleshooting of fault causes.

[0059] Knowledge graph data of the relationship between software: The relationship between the various software modules in the civil aviation passenger service system helps to quickly locate the software modules that may cause failures.

[0060] The above three types of knowledge are uploaded and updated in CSV, Word or PDF file format through the fault SMS knowledge maintenance module.

[0061] 2. Knowledge data preprocessing module:

[0062] This module is mainly responsible for preprocessing the above knowledge data, including data cleaning (removing duplicate data), chunking, etc. By dividing long text data into smaller processable units, the efficiency and accuracy of subsequent processing are improved.

[0063] 3. Data vector module:

[0064] This module uses the Langchain architecture to call the Embedding model (m3e-base) to convert the preprocessed knowledge data into vector representation. The vectorized data is stored in the vector database.

[0065] 4. Vector database:

[0066] The Faiss vector database is used to store and manage these vector data for efficient similarity calculation and retrieval.

[0067] 5. Vector retrieval module:

[0068] When a new fault message is received, the module uses the Langchain architecture to use the Embedding model (m3e-base) and the Rerank model (bge-reranker-v2-m3) to retrieve similar historical fault cases from the Faiss vector database. The most relevant fault-related nodes and solution-related data are quickly found through vector similarity calculation and result re-ranking.

[0069] 6.Large language model base:

[0070] The ChatGLM3-6B model is used as the core language processing base of the system. The model has powerful natural language understanding and generation capabilities, can accurately parse the semantics of fault text messages, and provide accurate fault analysis and suggestions in combination with knowledge graph data.

[0071] Through the collaborative work of the above modules, the present invention can realize intelligent analysis of fault text messages to quickly locate faults and provide relevant solutions, thereby improving the efficiency and accuracy of system fault handling.

[0072] Implementation principle

[0073] The present invention proposes a method for semantic parsing of fault SMS based on a civil aviation passenger service system, and the specific principle is as follows:

[0074] like Figure 2 As shown in FIG. 1 , the semantic parsing method of the fault SMS of the civil aviation passenger service system is specifically divided into the following steps:

[0075] Fault SMS analysis process:

[0076] Step 1: The interface monitors the content of the fault SMS.

[0077] Step 2: Perform a vector database search on the monitored fault SMS content to obtain the vector data with the highest correlation.

[0078] Step 3: Use the Rerank model to re-rank the search results in step 2 and select the top 5 ranking results.

[0079] Step 4: Assemble the content obtained in step 3 into prompt, for example:

[0080] prompt=f"""\

[0081] The historical fault data information is as follows:

[0082] ----------

[0083] {Step 3 Result}

[0084] ----------

[0085] Based on the historical fault data information, please analyze the following fault SMS content, the software that may have failed, and provide reference fault solutions. The output format should be:

[0086] Possible faulty software: <faulty software name>

[0087] Troubleshooting solutions (reference): <Troubleshooting solutions>

[0088] Fault SMS content: xxxxxxxx

[0089] """

[0090] Step 5: Call the large language model base with the prompt obtained in step 4 to get the final result.

[0091] Fault SMS knowledge maintenance process:

[0092] Step 1: Upload knowledge documents such as fault SMS history data, software responsibilities, and knowledge graph data of the relationship between software through the fault SMS knowledge maintenance module.

[0093] Step 2: Perform text extraction and text segmentation on the uploaded knowledge documents.

[0094] Step 3: Use the Langchain architecture to call the Embedding model (m3e-base) to vectorize the text processed in step 2.

[0095] Step 4: Store the vector-processed data in step 3 into the Faiss vector database for query by the vector retrieval module.

[0096] Fault SMS semantic parsing device based on civil aviation passenger service system Fault SMS parsing method:

[0097] 1. Upload knowledge files:

[0098] Upload knowledge documents such as fault SMS historical data, software responsibility files, and knowledge graph data on the relationship between software through the fault SMS knowledge maintenance module.

[0099] 2. Knowledge document preprocessing:

[0100] Generally, the historical fault SMS data is a csv file, including the content of the historical fault SMS and the solution. The software responsibilities are a word file, which mainly describes the functions and architecture of the civil aviation passenger service system software. The knowledge graph data of the relationship between software is a pdf file, which contains the upstream and downstream relationship between the various software in the civil aviation passenger service system.

[0101] Use Python's third-party libraries (python-docx, pdfplumber, pandas) to perform text extraction and text segmentation on the uploaded knowledge documents, and divide long documents into smaller text blocks for subsequent processing. The segmentation parameters are 100 characters per block, with 20 characters overlapping above and below.

[0102] 3. Data vector processing:

[0103] The Embedding model (m3e-base) is called using the Langchain architecture to convert the preprocessed text blocks into vector representations.

[0104] 4. Vector database storage:

[0105] The vectorized data is stored in the Faiss vector database for subsequent search by the vector retrieval module.

[0106] 5. Fault SMS monitoring:

[0107] After the system starts running, the fault SMS monitoring module receives and monitors the fault SMS content in the civil aviation passenger service system in real time.

[0108] 6. Vector search:

[0109] After receiving the fault SMS, the vector retrieval module is used to retrieve the content of the fault SMS.

[0110] The Langchain architecture is used to call the pre-trained Embedding model (m3e-base) to convert the fault SMS content into vectors, and the most relevant historical fault information is searched in the Faiss vector database. The retrieved data comes from the fault SMS related knowledge base, which contains the knowledge graph data of historical fault SMS data, software responsibilities and the relationship between software.

[0111] 7. Check the result sorting:

[0112] The Rerank model (bge-reranker-v2-m3) is used to re-rank the retrieved candidate results and obtain the top five most relevant fault information. This information may include fault type, related software modules, and historical solutions.

[0113] 8.Prompt assembly:

[0114] According to the reordered results, assemble and generate Prompt in the following format:

[0115] prompt=f"""\

[0116] The historical fault data points to the following:

[0117] ————————

[0118] {Rerank top five data}

[0119] ————————

[0120] Based on the historical fault data, please analyze the possible faults in the following fault SMS content and provide reference fault solutions. The output format should be:

[0121] Possible faulty software: <faulty software name>

[0122] Troubleshooting solutions (reference): <Troubleshooting solutions>

[0123] Fault SMS content: {Received fault SMS content}

[0124] """

[0125] 9. Calling the large language model:

[0126] Using the Prompt generated in the previous step, call the large language model (ChatGLM3-6B) to perform semantic analysis on the fault SMS and generate the final result, including the name of the software that may fail and the corresponding solution suggestions.

[0127] The above steps are the whole process of the device to analyze the fault text messages of the civil aviation passenger service system.

[0128] Embodiment 1: Fault SMS Knowledge Maintenance Process Example of Civil Aviation Passenger Service System:

[0129] Civil Aviation Passenger Service System Fault SMS Knowledge Files include:

[0130] 1. "Historical fault SMS data.csv"

[0131] As shown in Table 1, different troubleshooting solutions are given according to different fault SMS IDs.

[0132]

[0133]

[0134] Table 1 Contents of historical fault SMS data Use the fault SMS knowledge maintenance module to upload the "Historical fault SMS data.csv" file.

[0135] like Figure 3 As shown in the figure, the knowledge data preprocessing module uses the python third-party library pandas to read the content of "Historical Fault SMS Data.csv" and concatenate each line:

[0136] Fault SMS content: xxxx, troubleshooting solution: xxxx

[0137] Fault SMS content: xxxx, troubleshooting solution: xxxx

[0138] Put the processed content into a list, such as docs=[], and remove duplicates from the list.

[0139] If it is a word file or pdf file, use python to extract the text in the file, use chunk to divide every 100 characters into an element (the previous and next elements overlap by 20 characters), and then put them into a list. Use the Langchain framework to call the Embedding model (m3e-base) to store the data vector in the FAISS database.

[0140] FAISS.from_documents(docs,embeddings)

[0141] Note: embeddings = call of Embedding model (m3e-base).

[0142] Example 2: Fault SMS analysis process of civil aviation passenger service system:

[0143] Fault SMS:

[0144] Fault ID: 199005

[0145] Report a fault! 10:08 Receive reports from users

[0146] ______________

[0147] Time / Location: 10:06

[0148] Product / Software:xxxx

[0149] -----------------------

[0150] Fault phenomenon: Flight information cannot be updated

[0151] (1) Fault SMS monitoring: Receive the fault SMS "*****Fault phenomenon: Flight information cannot be updated". (2) Vector retrieval: Use the Langchain framework to call the Embedding model (m3e-base) to vectorize the SMS content, and retrieve the most relevant historical fault records in the Faiss vector database to take the first 256.

[0152] Example: [“Content of fault message: xxxx, troubleshooting solution: xxxx”, “Content of fault message: xxxx, troubleshooting solution: xxxx”, … “The flight query system is mainly responsible for flight information query, and flight data comes from the flight data synchronization module”]

[0153] (3) Inspection result ranking: Use the Rerank model (bge-reranker-v2-m3) to rerank the vector retrieval results. Keep the top five and determine the most relevant historical fault information.

[0154] (4) Prompt assembly: Generate a prompt based on the reordering result. The content is as follows: prompt = f"""\

[0155] The historical fault data points to the following:

[0156] ————————

[0157] 1. Fault SMS content: xxxx, troubleshooting and treatment plan: xxxx.

[0158] 2. Fault SMS content: xxxx, troubleshooting and treatment plan: xxxx.

[0159] …

[0160] 5. The flight query system is mainly responsible for flight information query, and the flight data comes from the flight data synchronization module.

[0161] ————————Please analyze the possible faults in the following fault SMS content based on the historical fault data content and provide reference fault solutions. The output format should be:

[0162] Possible faulty software: <faulty software name>

[0163] Troubleshooting solutions (reference): <Troubleshooting solutions>

[0164] Fault SMS content:

[0165] Fault ID: 199005

[0166] Report a fault! 10:08 Receive reports from users

[0167] ______________

[0168] Time / Location: 10:06

[0169] Product / Software:xxxx

[0170] -----------------------

[0171] Fault phenomenon: Flight information cannot be updated

[0172] """

[0173] (5) Call the large language model: Input Prompt into the large language model and obtain the parsing result:

[0174] ____________________________________________________

[0175] Software that may fail: Flight data synchronization module

[0176] Troubleshooting solution (reference): Restart the flight data synchronization service

[0177] ____________________________________________________

[0178] Return this as the final answer.

[0179] The present invention proposes an innovative solution to the problems of great difficulty in analyzing fault SMS and low efficiency in fault location in the prior art, and provides strong technical support for the stable operation of the civil aviation passenger service system.

[0180] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing one or more programs, which can implement the aforementioned method when the one or more programs are executed.

[0181] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A semantic parsing method for fault SMS based on a multilingual dialogue model, characterized in that: include: A fault SMS knowledge data storage step, storing and storing previous fault SMS knowledge data in the system; The fault SMS knowledge data preprocessing step preprocesses the fault SMS knowledge data, including data cleaning and block operations, so as to divide the long text data into smaller processable units, thereby improving the efficiency and accuracy of subsequent processing; In the data vector conversion step, the Embedding model is called using the Langchain architecture to convert the preprocessed knowledge data into vector representation, and the vectorized data is stored in the vector database. The Faiss vector database is used to store and manage these vector data. In the data vector retrieval step, the Embedding model and the Rerank model are called to retrieve similar historical fault cases from the Faiss vector database. The fault-related nodes and related data with the highest correlation are quickly found through vector similarity calculation and result re-ranking.

2. The semantic parsing method for fault SMS according to claim 1 is characterized in that: Further comprising the steps of: The steps for uploading fault SMS knowledge include uploading the knowledge graph data of fault SMS history data, software responsibilities, and the relationship between software.

3. The semantic parsing method for fault SMS according to claim 2 is characterized in that: Further comprising the steps of: Fault SMS knowledge data monitoring step: The interface monitors the fault SMS content and sends the monitored content to the fault SMS knowledge uploading step.

4. The semantic parsing method for fault SMS according to claim 2 is characterized in that: The fault SMS history data includes the content of historical fault SMS and processing solutions, the software responsibilities mainly describe the software functions and architecture of the civil aviation passenger service system, and the knowledge graph data of the relationship between software includes the upstream and downstream relationships between the various software in the civil aviation passenger service system.

5. The method for semantic parsing of fault SMS according to claim 4, characterized in that: The fault SMS knowledge data preprocessing step further includes: using Python's third-party libraries python-docx, pdfplumber and pandas to divide the above long text data into smaller text blocks for subsequent processing, and the cutting parameters are 100 characters per block, with 20 characters overlapping above and below.

6. The method for semantic parsing of fault SMS according to claim 5, characterized in that: The data vector retrieval step further includes: using a Rerank model to re-rank the retrieved candidate results by relevance, and obtaining the top five most relevant fault information, wherein the most relevant fault information includes fault type, relevant software modules and historical solutions.

7. The semantic parsing method for fault SMS according to claim 6 is characterized in that: The relevance reordering further includes: assembling and generating a Prompt according to the reordering result, wherein the Prompt includes: the name of the faulty software, the fault solution and the content of the received fault SMS message.

8. The method for semantic analysis of fault SMS according to claim 6, characterized in that: The Prompt is generated, the multi-language dialogue model is ChatGLM3-6B, ChatGLM3-6B is called, the fault SMS is semantically parsed and a final result is generated, the final result includes the name of the software that may fail and the corresponding solution suggestion.

9. A fault SMS semantic parsing device based on a multi-language dialogue model, characterized in that: include: Fault SMS knowledge data storage module, which stores and manages the previous fault SMS knowledge data in the system; The fault SMS knowledge data preprocessing module preprocesses the fault SMS knowledge data, including data cleaning and block operations, dividing the long text data into smaller processable units to improve the efficiency and accuracy of subsequent processing; The data vector conversion module uses the Langchain architecture to call the Embedding model, converts the preprocessed knowledge data into vector representation, stores the vectorized data into the vector database, and uses the Faiss vector database to store and manage these vector data; The data vector retrieval module calls the Embedding model and the Rerank model to retrieve similar historical fault cases from the Faiss vector database, and quickly finds the most relevant fault-related nodes and related data through vector similarity calculation and result re-ranking.

10. A computer-readable storage medium storing one or more programs, characterized in that: When the one or more programs are executed, the fault SMS semantic parsing method described in any one of claims 1 to 8 can be implemented.

Images