Language question and answer processing method and device for computer hardware management information

Abstract

The invention provides a language question and answer processing method and device for computer hardware management information. The method comprises the steps that first question information which is input by a user side and corresponds to the computer hardware management information is obtained; performing multi-thread asynchronous vectorization processing on the first problem information based on at least one working thread created in the thread pool to obtain a first problem vector, and performing retrieval processing in a preset embedded vector database to obtain a problem information retrieval result which is returned by the embedded vector database and is associated with the first problem vector; inputting the question information retrieval result and the first question information into a language question-answer processing model, and obtaining question reply information which is output by the language question-answer processing model and corresponds to the question information retrieval result and the first question information; the language question and answer processing model is obtained by training based on sample question information and reply label information. According to the method provided by the invention, the language question and answer processing efficiency and accuracy about the information of the baseboard management controller can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and device for processing language questions and answers of computer hardware management information, and also to an electronic device, a non-transitory computer-readable storage medium and a computer program product. Background Art

[0002] In recent years, with the continuous development of cloud computing and big data technology, computer hardware such as servers are increasingly widely used in various fields as key infrastructure, and their management and maintenance are becoming more and more important. At the same time, with the rapid development of computer technology, the management of computer hardware such as servers has become increasingly complex. Baseboard Management Controller (BMC) is a dedicated controller used to monitor and manage computer hardware such as servers. It is responsible for monitoring and managing the hardware status of computer hardware such as servers, reporting faults, etc. The existing operation mode of baseboard management controller requires users to have professional knowledge and skills, as well as certain experience, and the operation process is cumbersome. For R&D and testing personnel with less experience, they may not have a deep understanding of the baseboard management controller. In the actual operation process, they need to consult relevant manuals and materials to understand and operate the various computer hardware management information of the baseboard management controller, resulting in low efficiency. Therefore, how to design a more efficient operation query solution for computer hardware management information has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides a language question-answering processing method for computer hardware management information, which is used to solve the defect that the operation query scheme for computer hardware management information in the prior art has high limitations and leads to poor management efficiency.

[0004] The present invention provides a language question-answering processing method for computer hardware management information, comprising: Obtaining first question information corresponding to computer hardware management information input by a user terminal; Based on at least one worker thread created in the thread pool, multi-thread asynchronous vectorization processing is performed on the first question information to obtain a first question vector; based on the first question vector, a search process is performed in a preset embedded vector database to obtain a question information search result associated with the first question vector returned by the embedded vector database; The question information retrieval result and the first question information are input into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0005] According to a method for processing computer hardware management information language questions and answers provided by the present invention, at least one worker thread created in a thread pool performs multi-thread asynchronous vectorization processing on the first question information to obtain a first question vector; based on the first question vector, a search process is performed in a preset embedded vector database to obtain a question information search result associated with the first question vector returned by the embedded vector database, specifically including: Sending the first question information to the task queue of the thread pool, and calling the thread pool manager in the thread pool to create at least one corresponding working thread according to the number of the first question information in the task queue; wherein the task performed by each working thread in the at least one working thread is to perform vectorized processing on the corresponding first question information; Based on the at least one working thread, load the word embedding vector model corresponding to the at least one working thread respectively, and use the word embedding vector model to perform asynchronous vectorization processing on the first question information in the task queue to obtain a first question vector corresponding to the first question information; Based on the at least one working thread, the query interface corresponding to the embedded vector database is called respectively, and the first question vector is input into the embedded vector database through the query interface for retrieval and matching analysis, so as to obtain a first question information list associated with the first question vector returned by the embedded vector database; wherein the first question information list includes at least one question information arranged in order from front to back according to the similarity with the first question vector; wherein the embedded vector database pre-selects and stores multiple types of question information to be retrieved; The first question information list obtains the target number of question information as the question information retrieval result associated with the first question vector.

[0006] According to a method for processing computer hardware management information language questions and answers provided by the present invention, the step of obtaining first question information corresponding to the computer hardware management information input by a user terminal specifically includes: Obtaining initial question information corresponding to computer hardware management information input by the user; Inputting the initial question information into a language completeness detection model preset in a natural language processing library, performing completeness analysis on the initial question information using the language completeness detection model, and returning a list of second question information having a similar or identical meaning to the initial question information to the user terminal when it is determined that the language description of the initial question information is incomplete; The language integrity detection model is a neural network model trained based on sample initial question information and integrity label information corresponding to the sample initial question information; At least one target question information selected by the user terminal is determined based on the second question information list, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0007] According to a method for processing computer hardware management information through language question and answer, the first question vector is input into the embedded vector database for search and matching analysis, and a first question information list associated with the first question vector returned by the embedded vector database is obtained, which specifically includes: Using the first question vector as an index, respectively performing matching analysis with a plurality of types of question information to be retrieved that are pre-stored in the embedding vector database, so as to find at least one question information whose similarity with the first question vector reaches a preset similarity threshold from the embedding vector database; The at least one question information is arranged in order from front to back according to the similarity with the first question vector, so as to obtain a first question information list associated with the first question vector.

[0008] According to a language question-answering processing method for computer hardware management information provided by the present invention, the method of determining at least one target question information selected by the user terminal based on the second question information list, and using the at least one target question information as the first question information corresponding to the computer hardware management information specifically includes: Obtaining a selection operation request triggered by the user terminal based on the second question information list; At least one target question information selected by the user terminal from the second question information list is determined according to the selection operation request, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0009] According to a method for processing computer hardware management information through language question and answer provided by the present invention, before inputting the question information retrieval result and the first question information into a language question and answer processing model, the method further comprises: Obtaining sample question information for model training and answer label information corresponding to the sample question information; Iteratively training the initial language question and answer processing model based on the sample question information and the answer label information corresponding to the sample question information to obtain a trained language question and answer processing model; The network layers in the trained language question and answer processing model are pruned accordingly to obtain the language question and answer processing model.

[0010] The present invention also provides a language question-answering processing device for computer hardware management information, comprising: A problem information obtaining unit, used to obtain first problem information corresponding to the computer hardware management information input by the user terminal; a question information processing unit, configured to perform multi-thread asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool to obtain a first question vector; perform search processing on a preset embedded vector database based on the first question vector to obtain a question information search result associated with the first question vector returned by the embedded vector database; A question information answering unit is used to input the question information retrieval result and the first question information into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0011] According to a computer hardware management information language question and answer processing device provided by the present invention, the question information processing unit is specifically used for: Sending the first question information to the task queue of the thread pool, and calling the thread pool manager in the thread pool to create at least one corresponding working thread according to the number of the first question information in the task queue; wherein the task performed by each working thread in the at least one working thread is to perform vectorized processing on the corresponding first question information; Based on the at least one working thread, load the word embedding vector model corresponding to the at least one working thread respectively, and use the word embedding vector model to perform asynchronous vectorization processing on the first question information in the task queue to obtain a first question vector corresponding to the first question information; Based on the at least one working thread, the query interface corresponding to the embedded vector database is called respectively, and the first question vector is input into the embedded vector database through the query interface for retrieval and matching analysis, so as to obtain a first question information list associated with the first question vector returned by the embedded vector database; wherein the first question information list includes at least one question information arranged in order from front to back according to the similarity with the first question vector; wherein the embedded vector database pre-selects and stores multiple types of question information to be retrieved; The first question information list obtains the target number of question information as the question information retrieval result associated with the first question vector.

[0012] According to a computer hardware management information language question and answer processing device provided by the present invention, the question information obtaining unit is specifically used to: Obtaining initial question information corresponding to computer hardware management information input by the user; Inputting the initial question information into a language completeness detection model preset in a natural language processing library, performing completeness analysis on the initial question information using the language completeness detection model, and returning a list of second question information having a similar or identical meaning to the initial question information to the user terminal when it is determined that the language description of the initial question information is incomplete; The language integrity detection model is a neural network model trained based on sample initial question information and integrity label information corresponding to the sample initial question information; At least one target question information selected by the user terminal is determined based on the second question information list, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0013] According to a computer hardware management information language question and answer processing device provided by the present invention, the first question vector is input into the embedded vector database for search and matching analysis, and a first question information list associated with the first question vector returned by the embedded vector database is obtained, specifically comprising: Using the first question vector as an index, respectively performing matching analysis with a plurality of types of question information to be retrieved that are pre-stored in the embedding vector database, so as to find at least one question information whose similarity with the first question vector reaches a preset similarity threshold from the embedding vector database; The at least one question information is arranged in order from front to back according to the similarity with the first question vector, so as to obtain a first question information list associated with the first question vector.

[0014] According to a computer hardware management information language question and answer processing device provided by the present invention, the determining of at least one target question information selected by the user terminal based on the second question information list, and using the at least one target question information as the first question information corresponding to the computer hardware management information specifically includes: Obtaining a selection operation request triggered by the user terminal based on the second question information list; At least one target question information selected by the user terminal from the second question information list is determined according to the selection operation request, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0015] According to a computer hardware management information language question and answer processing device provided by the present invention, before inputting the question information retrieval result and the first question information into the language question and answer processing model, it also includes: The language question answering processing model acquisition unit is specifically used for: Obtaining sample question information for model training and answer label information corresponding to the sample question information; Iteratively training the initial language question and answer processing model based on the sample question information and the answer label information corresponding to the sample question information to obtain a trained language question and answer processing model; The network layers in the trained language question and answer processing model are pruned accordingly to obtain the language question and answer processing model.

[0016] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the language question-and-answer processing method for computer hardware management information as described in any one of the above items is implemented.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the language question-and-answer processing method for computer hardware management information as described in any of the above items is implemented.

[0018] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the language question-and-answer processing method for computer hardware management information as described in any one of the above items.

[0019] The language question and answer processing method for computer hardware management information provided by the present invention obtains first question information corresponding to the computer hardware management information input by the user end, performs multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool, obtains a first question vector, and performs retrieval processing on a preset embedded vector database based on the first question vector to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database; the question information retrieval result and the first question information are input into a language question and answer processing model to obtain output question answer information corresponding to the question information retrieval result and the first question information, which can effectively improve the efficiency and accuracy of language question and answer processing on baseboard management controller information, thereby enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0021] Figure 1 This is one of the flow charts of the language question-answering processing method for computer hardware management information provided by the present invention.

[0022] Figure 2 This is the second flow chart of the language question-and-answer processing method for computer hardware management information provided by the present invention.

[0023] Figure 3 This is the third flow chart of the language question-and-answer processing method for computer hardware management information provided by the present invention.

[0024] Figure 4 It is a structural schematic diagram of the language question and answer processing device for computer hardware management information provided by the present invention.

[0025] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0027] At present, the existing Baseboard Management Controller (BMC) management system can operate and manage server hardware through ipmi, redfish protocol and web pages. However, as the functions of the BMC management system become more and more powerful, the ipmi command management and Redfish protocol will become very numerous. At the same time, the various functions of the BMC management system become more complex, the hardware status it monitors will also become more complex, and some log collection and fault diagnosis knowledge will also increase. Therefore, if you use ipmi commands or Redfish to manage a powerful and complex BMC management system, you need to query the manual to find the corresponding protocol commands and usage methods, which is inefficient. At the same time, sometimes you may not know how to deal with some common fault information, such as what the different colors of the hard disk light mean or what some fault diagnosis logs mean. Even some experienced engineers may not be able to master all situations. In some cases, they may need to check the information or even check the relevant logs to analyze the reasons. It is time-consuming to directly check the information. With the continuous development and progress of artificial intelligence and natural language processing technologies, these technologies have begun to be widely used in various fields to improve workflows and user experience. Therefore, in order to improve the management and operation efficiency of the baseboard management controller, it is necessary to apply natural language processing technology to the baseboard management controller management system to meet the needs of server hardware management. In order to improve the management and operation efficiency of the BMC management system, the present invention applies natural language processing technology to the BMC management system, and combines the embedded vector database (ChromaDB) with the thread pool technology to improve the retrieval efficiency while ensuring the retrieval accuracy, and develops a language question-answering processing solution for computer hardware management information based on the embedded vector database to meet the needs of server hardware management.

[0028] Combine the following Figure 1-Figure 5 The present invention describes a method and device for processing computer hardware management information through language question and answer, and describes its embodiments in detail.

[0029] The following is a detailed description of an embodiment of the method for processing computer hardware management information through language question and answer based on the present invention. Figure 1 As shown, it is one of the flow charts of the language question-answering processing method of computer hardware management information provided by the present invention, and the specific implementation process includes the following steps: Step 101: Obtain first question information corresponding to computer hardware management information input by a user terminal.

[0030] In an embodiment of the present invention, it is first necessary to obtain the initial question information corresponding to the computer hardware management information input by the user terminal through the user interface input part, and then input the initial question information into the language completeness detection model preset in the natural language processing library, and use the language completeness detection model to perform completeness analysis on the initial question information. When it is determined that the language description of the initial question information is incomplete, a second question information list with a similar or identical meaning to the initial question information is returned to the user terminal. Furthermore, based on the second question information list, at least one target question information selected by the user terminal is determined, and the at least one target question information is used as the first question information corresponding to the computer hardware management information. Among them, the language completeness detection model is a neural network model trained based on sample initial question information and the completeness label information corresponding to the sample initial question information. Specifically, in the process of determining at least one target question information selected by the user terminal based on the second question information list, and using the at least one target question information as the first question information corresponding to the computer hardware management information, a selection operation request triggered by the user terminal based on the second question information list can be obtained, and at least one target question information selected by the user terminal from the second question information list can be determined according to the selection operation request, and the at least one target question information can be used as the first question information corresponding to the computer hardware management information.

[0031] It should be noted that the user interface input part mainly shows the question and answer module implemented in the web interface of the BMC management system. When the user clicks on the question and answer module, a corresponding chat session interface is generated. The user can have a language conversation in the chat session interface, and can also ask questions about the BMC management system or computer hardware management information such as servers, such as: relevant knowledge and common functions of the BMC management system, how to operate and manage the BMC, the use of the common protocol Redfish and IPMI commands, and knowledge related to BMC fault diagnosis, etc., all of which can be asked from the chat session interface. The chat session interface has a built-in NLP.js library (i.e., natural language processing library), and a trained lightweight model (i.e., language integrity detection model) is pre-stored in the natural language processing library. The language integrity detection model will determine whether the initial question information input by the user end is complete. When the initial question information input by the user end is detected to be incomplete, at least one expression with similar meaning will be returned for the user to select (i.e., a second question information list with similar or identical meaning to the initial question information will be returned to the user end). For example, when the user end enters "shutdown", the content of the second question information list returned by the chat session interface may include: you want to know "how to shut down through the ipmitool command", "how to shut down through redfish", and "how to shut down through the web interface". When the user end selects the target question information to be asked, the target question information is transferred to the embedded vector database.

[0032] Step 102: Based on at least one working thread created in the thread pool, perform multi-threaded asynchronous vectorization processing on the first question information to obtain a first question vector; based on the first question vector, perform retrieval processing on a preset embedded vector database to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database.

[0033] In an embodiment of the present invention, the first question information is sent to the task queue of the thread pool, and the thread pool manager in the thread pool is called to create at least one corresponding worker thread according to the number of the first question information in the task queue. The task performed by each worker thread in the at least one worker thread is to vectorize the corresponding first question information. Based on the at least one worker thread, the word embedding vector model corresponding to the at least one worker thread is loaded, and the word embedding vector model is used to asynchronously vectorize the first question information in the task queue to obtain a first question vector corresponding to the first question information (i.e., Figure 3Vector 1, vector 2, etc. generated as shown). Based on the at least one working thread, the query interface corresponding to the embedded vector database is called respectively, and the first question vector is input into the embedded vector database through the query interface for retrieval and matching analysis, so as to obtain a first question information list associated with the first question vector returned by the embedded vector database; wherein the first question information list contains at least one question information arranged in sequence from front to back according to the similarity with the first question vector. Among them, the embedded vector database pre-selects and stores multiple types of question information to be retrieved. The first question information list obtains the target number of question information as the question information retrieval result associated with the first question vector. Figure 2 As shown: When the input question (i.e., the first question information) can be "How to collect one-click logs", the search results matched by the embedded vector database (i.e., the question information search results) can be: (1) One-click logs can be collected through redfish. (2) One-click logs can be collected through the web page: Enter device_ip in the web browser to log in, click the 'log' tab on the left sidebar, and then click the log collection tab in the expanded tab. The log collection operation interface will appear on the right. Click the log collection page on the right to collect logs. (3) Use redfish to view the log collection status: The answer output is obtained by analyzing the search results + questions through the NLP model: Hello, there are currently two ways to collect one-click logs.

[0034] In the process of inputting the first question vector into the embedded vector database for search and matching analysis, and obtaining the first question information list associated with the first question vector returned by the embedded vector database, the first question vector needs to be used as an index in advance, and matching analysis is performed with the plurality of types of question information to be retrieved stored in the embedded vector database respectively, so as to find at least one question information whose similarity with the first question vector reaches a preset similarity threshold from the embedded vector database. The at least one question information is arranged in order from front to back according to the similarity with the first question vector, and the first question information list associated with the first question vector is obtained.

[0035] It should be noted that the embedded vector database is a lightweight vector database ChromaDB, which is an open source embedded vector database for storing and retrieving vector embeddings. The embedded vector database can perform searches asynchronously, which improves the retrieval speed. Before performing the search, the first question information needs to be vectorized through the word embedding vector model (all-MiniLM-L6-v2) preset in the embedded vector database, and the first question information is vectorized through the word embedding vector model in the embedded vector database, and the question text corresponding to the first question information is converted into a digital vector (i.e., the first question vector). For example, the first question information input is "How to shut down through the ipmitool command", and the word embedding vector model can convert it into [3.8173e-02; 4.1413e-02; 6.8678e-02; 8.8600e-03; 1.3012e-02...], thereby improving the search efficiency.

[0036] In addition, facing the huge number of users of the BMC management system, if each question text is vectorized in sequence, the retrieval speed will be reduced to a certain extent, resulting in a poor user experience. Although the embedded vector database has a cache mechanism that can store some common questions, the cache space is also limited. For the huge number of users of the BMC management system, the previous cache results will soon be overwritten by new question texts. To this end, the present invention provides a method of improving retrieval efficiency by combining a thread pool with ChromaDB. Specifically, when the user terminal inputs the question text corresponding to the first question information, the word embedding vector model is no longer directly loaded for quantization processing, but the task of loading the quantization processing of the word embedding vector model is placed in the working thread in the thread pool. When a large amount of question information (i.e., a large amount of first question information) is input, the thread pool manager (i.e., the management thread) in the thread pool will start the corresponding working thread to load the word embedding vector model, and use the corresponding algorithm to vectorize the corresponding question text. For example, when a large number of user terminals are currently using language question and answer processing, if there is no management thread in the thread pool to convert the question text input into a digital vector, then each user terminal question will become a linear operation, which will reduce the quantization speed to a certain extent, thereby affecting the speed of answering questions. To this end, the present invention uses a thread pool to asynchronously manage text quantization work. When there are many users using language question and answer processing, a large number of question texts corresponding to the first question information need to be quantized. The task interface in the thread pool adds the task to the task queue. The management thread will create a corresponding worker thread according to the number of tasks in the task queue (for example, the number of worker threads is 5). The task performed by each worker thread is to vectorize the corresponding question text. Then the worker thread calls the corresponding query API of the embedded vector database, and uses the vectorized question text (such as the vector form of [3.8173e-02; 4.1413e-02; 6.8678e-02; 8.8600e-03; 1.3012e-02...]) as input information. The embedded vector database returns the search result, and specifically, the question information of the most similar front target number (for example, the first 3) can be retrieved as the question information retrieval result associated with the first question vector. Then the question information retrieval result and the input first question vector are used as the input of the language question and answer processing model, so that the language question and answer processing model can understand and generate answers. The thread pool is a form of multithreading processing used to manage and reuse threads to improve system performance and resource utilization. It uses pooling technology to reuse resources, avoid the overhead of frequent creation and destruction of threads, thereby reducing resource consumption and improving response speed. The thread pool consists of a thread pool manager, a task queue, and a worker thread. When the application starts, a certain number of threads are created and saved in the thread pool.When a task needs to be executed, an idle thread is obtained from the thread pool to assign the task. After the task is completed, the thread returns to the thread pool and waits to execute the next task.

[0037] Step 103: input the question information retrieval result and the first question information into a language question and answer processing model, and obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0038] In an embodiment of the present invention, the question information retrieval result and the first question information may be input into a language question answering processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question answering processing model. The question information retrieval result includes at least one question information similar to the first question information. The language question answering processing model is a neural network model trained based on sample question information and answer label information corresponding to the sample question information.

[0039] It should be noted that before inputting the question information retrieval result and the first question information into the language question and answer processing model, it is also necessary to obtain sample question information for model training and the answer label information corresponding to the sample question information in advance, perform iterative training on the initial language question and answer processing model based on the sample question information and the answer label information corresponding to the sample question information, obtain the trained language question and answer processing model, perform corresponding pruning on the network layers in the trained language question and answer processing model, and obtain the language question and answer processing model.

[0040] The language question and answer processing model is a natural language processing model (NLP). Due to the limited resources of the BMC management system, the existing language question and answer processing model cannot be used directly, and the language question and answer processing model needs to be trimmed. Specifically, a trimming processing method combining structured trimming and unstructured trimming can be adopted according to the performance and structural complexity of the language question and answer processing model. In the case where the question information retrieval result and the first question information to be processed are speech recognition tasks, the multiple speech data corresponding to the question information retrieval result and the first question information to be recognized can be input into the trimmed language question and answer processing model, and based on the convolution kernels of each network layer after the corresponding mixed trimming in the trimmed language question and answer processing model, the multiple speech data corresponding to the question information retrieval result and the first question information to be recognized are respectively subjected to speech recognition processing, and the recognition processing results of the corresponding speech data are obtained, and the question answer information corresponding to the question information retrieval result and the first question information is generated according to the recognition processing results. Among them, the convolution kernels of each network layer after the corresponding mixed clipping process are obtained by mixed clipping according to the characteristics of the speech recognition task (such as focusing on recognizing speech data) and the importance value of the corresponding network layer. The efficiency of loading and running the language question and answer processing model can be effectively improved by clipping the language question and answer processing model and then loading it. For example: in the clipping process, the network layer with less contribution can be removed (for example, from 24 network layers to 6 network layers); the number of parameters of the language question and answer processing model can be reduced by reducing the hidden layer dimension of each layer, adjusting the hidden layer dimension layer by layer and monitoring the performance change of the language question and answer processing model (for example, 786 dimensions are reduced to 384 dimensions); limiting the maximum support sequence length (for example, 8k is reduced to 1k), the maximum support sequence can be understood as how many tokens can be input at a time (for example, in the field of large prediction models, one token can represent 2-3 Chinese characters), and the clipped language question and answer processing model can be loaded and run inside the BMC management system after fine-tuning. The data set used to fine-tune the language question-answering processing model includes the BMC basic knowledge manual, the IPMI / Redfish protocol user manual, the fault diagnosis knowledge manual, and other BMC-related information, so that the lightweight language question-answering processing model can have the ability to generate corresponding BMC answers based on relevant question texts. After the first question information input by the user end is retrieved through the embedded vector database, the question information retrieval result and the first question information are input into the language question-answering processing model, which facilitates the generation and understanding of the answer by the language question-answering processing model, so as to obtain the question answer information corresponding to the question information retrieval result and the first question information.In addition, in order to enable the language question and answer processing model to be loaded and run within the BMC management system, the present invention also builds a Python environment for the operation of the language question and answer processing model within the BMC management system, and introduces relevant library files, such as (Pytorch, Transforms) and other library files that support the operation of large models, so that the language question and answer processing model can be loaded and run normally.

[0041] The language question and answer processing method for computer hardware management information of the present invention obtains first question information corresponding to the computer hardware management information input by the user end, performs multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool, obtains a first question vector, and performs retrieval processing on a preset embedded vector database based on the first question vector, obtains a question information retrieval result associated with the first question vector returned by the embedded vector database, inputs the question information retrieval result and the first question information into a language question and answer processing model, obtains question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model, and directly asks questions through an intelligent question and answer system, making the operation and maintenance of the BMC management system more efficient, and can effectively improve the efficiency and accuracy of language question and answer processing on baseboard management controller information, avoids the need to look up manuals to inquire about knowledge related to the BMC management system, thereby improving the user experience.

[0042] The following is a description of the computer hardware management information language question and answer processing device provided by the present invention. The computer hardware management information language question and answer processing device described below and the computer hardware management information language question and answer processing method described above can be referred to each other. Figure 4 As shown, it is a schematic diagram of the structure of the language question and answer processing device for computer hardware management information provided by the present invention. The language question and answer processing device for computer hardware management information of the present invention specifically includes the following parts: The problem information obtaining unit 401 is used to obtain first problem information input by a user terminal and corresponding to the computer hardware management information.

[0043] The question information processing unit 402 is used to perform multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool to obtain a first question vector; perform retrieval processing on a preset embedded vector database based on the first question vector to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database.

[0044] The question information answering unit 403 is used to input the question information retrieval result and the first question information into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0045] According to a computer hardware management information language question and answer processing device provided by the present invention, the question information processing unit is specifically used for: Sending the first question information to the task queue of the thread pool, and calling the thread pool manager in the thread pool to create at least one corresponding working thread according to the number of the first question information in the task queue; wherein the task performed by each working thread in the at least one working thread is to perform vectorized processing on the corresponding first question information; Based on the at least one working thread, load the word embedding vector model corresponding to the at least one working thread respectively, and use the word embedding vector model to perform asynchronous vectorization processing on the first question information in the task queue to obtain a first question vector corresponding to the first question information; Based on the at least one working thread, the query interface corresponding to the embedded vector database is called respectively, and the first question vector is input into the embedded vector database through the query interface for retrieval and matching analysis, so as to obtain a first question information list associated with the first question vector returned by the embedded vector database; wherein the first question information list includes at least one question information arranged in order from front to back according to the similarity with the first question vector; wherein the embedded vector database pre-selects and stores multiple types of question information to be retrieved; The first question information list obtains the target number of question information as the question information retrieval result associated with the first question vector.

[0046] According to a computer hardware management information language question and answer processing device provided by the present invention, the question information obtaining unit is specifically used to: Obtaining initial question information corresponding to computer hardware management information input by the user; Inputting the initial question information into a language completeness detection model preset in a natural language processing library, performing completeness analysis on the initial question information using the language completeness detection model, and returning a list of second question information having a similar or identical meaning to the initial question information to the user terminal when it is determined that the language description of the initial question information is incomplete; The language integrity detection model is a neural network model trained based on sample initial question information and integrity label information corresponding to the sample initial question information; At least one target question information selected by the user terminal is determined based on the second question information list, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0047] According to a computer hardware management information language question and answer processing device provided by the present invention, the first question vector is input into the embedded vector database for search and matching analysis, and a first question information list associated with the first question vector returned by the embedded vector database is obtained, specifically comprising: Using the first question vector as an index, respectively performing matching analysis with a plurality of types of question information to be retrieved that are pre-stored in the embedding vector database, so as to find at least one question information whose similarity with the first question vector reaches a preset similarity threshold from the embedding vector database; The at least one question information is arranged in order from front to back according to the similarity with the first question vector, so as to obtain a first question information list associated with the first question vector.

[0048] According to a computer hardware management information language question and answer processing device provided by the present invention, the determining of at least one target question information selected by the user terminal based on the second question information list, and using the at least one target question information as the first question information corresponding to the computer hardware management information specifically includes: Obtaining a selection operation request triggered by the user terminal based on the second question information list; At least one target question information selected by the user terminal from the second question information list is determined according to the selection operation request, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

[0049] According to a computer hardware management information language question and answer processing device provided by the present invention, before inputting the question information retrieval result and the first question information into the language question and answer processing model, it also includes: The language question answering processing model acquisition unit is specifically used for: Obtaining sample question information for model training and answer label information corresponding to the sample question information; Iteratively training the initial language question and answer processing model based on the sample question information and the answer label information corresponding to the sample question information to obtain a trained language question and answer processing model; The network layers in the trained language question and answer processing model are pruned accordingly to obtain the language question and answer processing model.

[0050] The language question and answer processing device for computer hardware management information provided by the present invention obtains first question information corresponding to the computer hardware management information input by the user end, performs multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool, obtains a first question vector, and performs retrieval processing on a preset embedded vector database based on the first question vector to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database; the question information retrieval result and the first question information are input into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model, which can effectively improve the efficiency and accuracy of language question and answer processing on baseboard management controller information, thereby enhancing the user experience.

[0051] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device (i.e., electronic device) may include: a processor 501, a communication interface 504, a memory 502, and a communication bus 503, wherein the processor 501, the communication interface 504, and the memory 502 communicate with each other through the communication bus 503. The processor 501 may call the logic instructions in the memory 502 to execute the language question and answer processing method of computer hardware management information, the method comprising: obtaining first question information corresponding to the computer hardware management information input by the user end; performing multi-thread asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool to obtain a first question vector; performing search processing on a preset embedded vector database based on the first question vector to obtain a question information search result associated with the first question vector returned by the embedded vector database; inputting the question information search result and the first question information into a language question and answer processing model to obtain question answer information corresponding to the question information search result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and the answer label information corresponding to the sample question information.

[0052] In addition, the logic instructions in the above-mentioned memory 502 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0053] On the other hand, the present application also provides a computer program product, which includes a computer program, which can be stored on a computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the language question and answer processing method of computer hardware management information provided by the above methods, the method including: obtaining first question information corresponding to the computer hardware management information input by a user end; performing multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in a thread pool to obtain a first question vector; performing retrieval processing on a preset embedded vector database based on the first question vector to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database; inputting the question information retrieval result and the first question information into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0054] On the other hand, the present application also provides a computer-readable storage medium, which includes a stored program, wherein the program, when running, executes the language question and answer processing method of computer hardware management information provided by the above methods, the method comprising: obtaining first question information corresponding to the computer hardware management information input by a user terminal; performing multi-threaded asynchronous vectorization processing on the first question information based on at least one working thread created in a thread pool to obtain a first question vector; performing retrieval processing on a preset embedded vector database based on the first question vector to obtain a question information retrieval result associated with the first question vector returned by the embedded vector database; inputting the question information retrieval result and the first question information into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

[0055] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0056] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application 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 replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for processing computer hardware management information by language question and answer, characterized in that: include: Obtaining first question information corresponding to computer hardware management information input by a user terminal; Based on at least one worker thread created in the thread pool, multi-thread asynchronous vectorization processing is performed on the first question information to obtain a first question vector; based on the first question vector, a search process is performed in a preset embedded vector database to obtain a question information search result associated with the first question vector returned by the embedded vector database; The question information retrieval result and the first question information are input into a language question and answer processing model to obtain question answer information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

2. The method for processing computer hardware management information by language question and answer according to claim 1, characterized in that: The at least one worker thread created in the thread pool performs multi-thread asynchronous vectorization processing on the first question information to obtain a first question vector; and performs search processing on a preset embedded vector database based on the first question vector to obtain a question information search result associated with the first question vector returned by the embedded vector database, specifically including: Sending the first question information to the task queue of the thread pool, and calling the thread pool manager in the thread pool to create at least one corresponding working thread according to the number of the first question information in the task queue; wherein the task performed by each working thread in the at least one working thread is to perform vectorized processing on the corresponding first question information; Based on the at least one working thread, load the word embedding vector model corresponding to the at least one working thread respectively, and use the word embedding vector model to perform asynchronous vectorization processing on the first question information in the task queue to obtain a first question vector corresponding to the first question information; Based on the at least one working thread, the query interface corresponding to the embedded vector database is called respectively, and the first question vector is input into the embedded vector database through the query interface for retrieval and matching analysis, so as to obtain a first question information list associated with the first question vector returned by the embedded vector database; wherein the first question information list includes at least one question information arranged in order from front to back according to the similarity with the first question vector; wherein the embedded vector database pre-selects and stores multiple types of question information to be retrieved; The first question information list obtains the target number of question information as the question information retrieval result associated with the first question vector.

3. The language question-answering processing method for computer hardware management information according to claim 1, characterized in that: The obtaining of the first question information corresponding to the computer hardware management information input by the user terminal specifically includes: Obtaining initial question information corresponding to computer hardware management information input by the user; Inputting the initial question information into a language completeness detection model preset in a natural language processing library, performing completeness analysis on the initial question information using the language completeness detection model, and returning a list of second question information having a similar or identical meaning to the initial question information to the user terminal when it is determined that the language description of the initial question information is incomplete; The language integrity detection model is a neural network model trained based on sample initial question information and integrity label information corresponding to the sample initial question information; At least one target question information selected by the user terminal is determined based on the second question information list, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

4. The method for processing computer hardware management information by language question and answer according to claim 2, characterized in that: The step of inputting the first question vector into the embedded vector database for search and matching analysis to obtain a first question information list associated with the first question vector returned by the embedded vector database specifically includes: Using the first question vector as an index, respectively performing matching analysis with a plurality of types of question information to be retrieved that are pre-stored in the embedding vector database, so as to find at least one question information whose similarity with the first question vector reaches a preset similarity threshold from the embedding vector database; The at least one question information is arranged in order from front to back according to the similarity with the first question vector, so as to obtain a first question information list associated with the first question vector.

5. The method for processing computer hardware management information by language question and answer according to claim 3, characterized in that: The determining at least one target question information selected by the user terminal based on the second question information list, and using the at least one target question information as the first question information corresponding to the computer hardware management information specifically includes: Obtaining a selection operation request triggered by the user terminal based on the second question information list; At least one target question information selected by the user terminal from the second question information list is determined according to the selection operation request, and the at least one target question information is used as the first question information corresponding to the computer hardware management information.

6. The method for processing computer hardware management information by language question and answer according to claim 1, characterized in that: Before inputting the question information retrieval result and the first question information into the language question answering processing model, the method further includes: Obtaining sample question information for model training and answer label information corresponding to the sample question information; Iteratively training the initial language question and answer processing model based on the sample question information and the answer label information corresponding to the sample question information to obtain a trained language question and answer processing model; The network layers in the trained language question and answer processing model are pruned accordingly to obtain the language question and answer processing model.

7. A computer hardware management information language question and answer processing device, characterized in that: include: A problem information obtaining unit, used to obtain first problem information corresponding to the computer hardware management information input by the user terminal; a question information processing unit, configured to perform multi-thread asynchronous vectorization processing on the first question information based on at least one working thread created in the thread pool to obtain a first question vector; perform search processing on a preset embedded vector database based on the first question vector to obtain a question information search result associated with the first question vector returned by the embedded vector database; A question information answering unit is used to input the question information retrieval result and the first question information into a language question and answer processing model to obtain question answering information corresponding to the question information retrieval result and the first question information output by the language question and answer processing model; wherein the language question and answer processing model is trained based on sample question information and answer label information corresponding to the sample question information.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the language question and answer processing method for computer hardware management information according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the language question-and-answer processing method for computer hardware management information according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the language question-and-answer processing method for computer hardware management information according to any one of claims 1 to 6 is implemented.

Images