Data processing method for power conversion operation and maintenance, computing device and readable storage medium

Abstract

The invention discloses a data processing method for power conversion operation and maintenance, computing equipment and a computer readable storage medium, and the method comprises the steps: obtaining a to-be-stored file for power conversion operation and maintenance, and determining the target description information of the to-be-stored file; splitting the target description information, and determining at least one first target text fragment; and according to the at least one first target text fragment, determining at least one index vector of the to-be-stored file, and representing the to-be-stored file based on the at least one index vector. Thus, the file is represented through the index vector, different types of data output can be generated more conveniently according to actual requirements, and meanwhile the accuracy of data output is improved.

Description

Technical Field

[0001] The present application relates to the field of large models, and in particular to a data processing method, computing device, and readable storage medium for battery swapping operation and maintenance. Background Art

[0002] With the increasing popularity of new energy vehicles, battery swap stations, as crucial infrastructure for electric vehicle energy replenishment, are expanding in number and scale. Managing battery swap operation and maintenance data poses a key challenge: accurately outputting the data and information users require and resolving issues. Currently, when processing or outputting data based on user questions, there are issues with inaccurate data output, limited data formatting, and the ability to only output text. Summary of the Invention

[0003] The purpose of this application is to provide a data processing method, computing device, and readable storage medium for battery swap operation and maintenance, which can more conveniently generate different types of data output according to actual needs, while improving the accuracy of data output.

[0004] To achieve the above objectives: In a first aspect, an embodiment of the present application provides a data processing method for battery swapping operation and maintenance, including: Obtain the files to be stored for battery swap operation and maintenance, and determine the target description information of the files to be stored; Splitting the target description information to determine at least one first target text segment; At least one index vector of the file to be stored is determined according to the at least one first target text segment, so as to represent the file to be stored based on the at least one index vector.

[0005] In one embodiment, the file to be stored includes image data, and obtaining the file to be stored for battery swapping operation and maintenance and determining target description information of the file to be stored includes: Determining first description information carried by the image data in text form; Acquire second description information of the context associated with the image data in the form of text; The first description information and the second description information are converted into a target format and determined as target description information of the file to be stored.

[0006] In one embodiment, splitting the target description information to determine at least one first target text segment includes: According to the preset battery swap operation and maintenance type, the target description information is split and processed to determine at least one initial text segment; The multiple initial text segments are further split according to the hierarchical information and / or the degree of association of contextual semantics to determine at least one first target text segment.

[0007] In one embodiment, determining at least one index vector of the file to be stored based on the at least one first target text segment includes: Determining an index vector corresponding to at least one first target text segment in the file to be stored; The index vector corresponding to the at least one first target text segment is determined as at least one index vector of the file to be stored.

[0008] In one embodiment, determining the index vector corresponding to at least one first target text segment in the file to be stored includes: Determining third description information that continues with the first target text segment; determining summary information of the first target text segment according to the third description information and the first target text segment; determining a query language of the first target text segment; An index vector of the first target text segment is determined according to the summary information and the query language.

[0009] In one embodiment, determining the query language of the first target text segment includes: A preset number of query languages related to the first target text segment are determined according to the first target text segment and the third description information.

[0010] In one embodiment, the method further comprises: Obtaining requests for battery swap operations and maintenance; Determining at least one index vector that matches the request question according to a preset matching method; Determining a corresponding second target text segment according to the at least one index vector; According to the second target text segment, the target file for battery swapping operation and maintenance is determined and obtained in the battery swapping operation and maintenance database.

[0011] In one embodiment, the method further comprises: Obtain the target file as the answer result, and evaluate the accuracy of the answer result based on the preset test data set, including: Determining the accuracy of the answer results based on the number of questions answered correctly and the total number of questions; and / or, In a multi-round dialogue, the accuracy of the answer result is determined based on the number of multi-round dialogues without errors and the total number of multi-round dialogues.

[0012] In a second aspect, an embodiment of the present application provides a computing device, specifically including: A processor; A memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the instructions for performing the data processing method of battery swapping operation and maintenance as described in the first aspect.

[0013] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the instructions in the computer-readable storage medium are executed by the processor of a computing device, the computing device can implement the data processing method of battery swapping operation and maintenance as described in the first aspect.

[0014] A data processing method, a computing device, and a computer-readable storage medium for battery swapping operation and maintenance provided by an embodiment of the present application include: obtaining a file to be stored for battery swapping operation and maintenance, and determining target description information of the file to be stored; splitting the target description information to determine at least one first target text segment; and determining at least one index vector of the file to be stored according to the at least one first target text segment, so as to represent the file to be stored based on the at least one index vector. In this way, by representing the file through the index vector, different types of data outputs can be generated more conveniently according to actual needs, and at the same time, the accuracy of the data output is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flow schematic of the data processing method for battery swapping operation and maintenance provided by an embodiment of the present invention Figure 1 ; Figure 2 Flow schematic of the data processing method for battery swapping operation and maintenance provided by an embodiment of the present invention Figure 2 ; Figure 3 Specific flow schematic of the data processing method for battery swapping operation and maintenance provided by an embodiment of the present invention; Figure 4 Specific flow schematic of processing a file to be stored provided by an embodiment of the present invention; Figure 5 Structure schematic of the computing device provided by an embodiment of the present invention; Processor 210, memory 211, network interface 212, bus system 213. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0017] It should be noted that in this document, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of additional identical elements in the process, method, article or device comprising such element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined based on their explanations in the specific embodiments or further in combination with the context of the specific embodiments.

[0018] It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this document, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining". Furthermore, as used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms unless the context indicates otherwise. It should be further understood that the terms "comprising", "including" indicate the presence of the stated features, steps, operations, elements, components, items, types, and / or groups, but do not exclude the presence, occurrence or addition of one or more other features, steps, operations, elements, components, items, types, and / or groups. The terms "or" and "and / or" used herein are interpreted as inclusive, or meaning any one or any combination. Thus, "A, B or C" or "A, B and / or C" means "any of the following: A; B; C; A and B; A and C; B and C; A, B and C". An exception to this definition occurs only when the combination of elements, functions, steps or operations is inherently mutually exclusive in some way.

[0019] It should be understood that although the steps in the flowchart in the embodiments of the present application are sequentially shown according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and they can be executed in other orders. Moreover, at least a part of the steps in the figure may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments, and their execution order is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0020] It should be noted that in this article, step codes such as S101 and S102 are used. The purpose is to more clearly and briefly express the corresponding content and do not constitute a substantial limitation in order. Those skilled in the art may execute S102 first and then S101 during specific implementation, etc., but these should all be within the protection scope of the present application.

[0021] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0022] In the following description, suffixes such as "module", "component" or "unit" used to represent elements are only for the convenience of explaining the present application, and they have no specific meaning in themselves. Therefore, "module", "component" or "unit" can be used interchangeably.

[0023] Refer to Figure 1 , the embodiments of the present application provide a data processing method for battery swapping operation and maintenance. This data processing method for battery swapping operation and maintenance can be implemented in software and / or hardware. In this embodiment, taking the data processing method for battery swapping operation and maintenance applied to a computing device as an example, the data processing method for battery swapping operation and maintenance provided in this embodiment includes: Step S101: Obtain the file to be stored for battery swapping operation and maintenance, and determine the target description information of the file to be stored.

[0024] Optionally, during the normal process of battery swapping operation and maintenance, various types of battery swapping operation and maintenance data generated during the battery swapping operation and maintenance process can be obtained and determined as the file to be stored, including operation data of battery swapping equipment, maintenance record operations, order data, and other information related to battery swapping operation and maintenance. Here, the file to be stored may include text content, picture content, video content, audio content, etc.

[0025] Optionally, by setting target description information, it is ensured that these files to be stored can be accurately identified and effectively utilized in the subsequent storage and management process. Specifically, the target description information can include basic attribute information of the file, such as file name, creation time, modification time, etc. At the same time, the target description information should be combined with the business logic and data dictionary of battery swap operation and maintenance to conduct an in-depth analysis of the file content to identify the specific business areas, operation links and related battery swap operation and maintenance information involved in the file. Here, the various information extracted and parsed can also be integrated and standardized according to the preset target description information template to form complete and accurate target description information, which provides a solid foundation for subsequent file storage, retrieval and data analysis.

[0026] In one embodiment, the file to be stored includes image data, obtaining the file to be stored for battery swap operation and maintenance, and determining target description information of the file to be stored include: Determining first description information carried by the image data in text form; Obtaining second description information of the context associated with the image data in the form of text; The first description information and the second description information are converted into a target format and determined as target description information of the file to be stored.

[0027] Optionally, the image data in the stored file is described in the form of text by loading text or through OCR recognition technology. The first description information specifically includes at least one of the following: file format information; features, subjects and object elements contained in the image data. For example, if it is a landscape picture, the category of the scenery, such as mountains, rivers, oceans, etc., is identified through image recognition technology, and then these key information is recorded in the form of text; the connection relationship / interrelationship between each feature in the image data. Optionally, while determining the first description information, the original data of the image data (i.e., the image source) is saved in the database of the exchange point operation and maintenance to ensure the integrity and traceability of the image data, so that the original image can be obtained at any time when needed, and it is also conducive to further management and analysis of the image data.

[0028] Optionally, when describing the context of the image data in the storage file in the form of text and obtaining the corresponding second description information, the location of the image data, the information contained in the context of the image data, and the relationship between the image data and the corresponding context can be described separately based on the technical method of loading text and OCR recognition, and integrated to generate the second description information. In this way, the second description information further explains the image data, which enables the large model to understand and determine the image data more accurately.

[0029] Optionally, after obtaining the first description information and the second description information, the first description information and the second description information are integrated and converted according to a preset target format. Here, the target format can be defined according to specific application scenarios and requirements, and can specifically be a specific database table structure, XML format, or other custom data structures. Here, it is generally set to the general format of markdown. By converting the first description information and the second description information into the target format, the resulting target description information can be made more standardized and normalized, which helps to better implement data management and facilitates transmission and sharing between different systems.

[0030] In other embodiments, for the text data contained in the file to be stored, all types of data information can be classified and labeled according to classification tags such as the operation and maintenance, repair, maintenance, and battery swapping process of the battery swapping station, and saved in different directories according to a preset hierarchical method. Further, all the text content in the file to be stored is recognized by loading the text and through OCR recognition, and then all the text content is processed by a large model, and the output result is in the general format of markdown, which is determined as the target description information of the file to be stored.

[0031] Step S102: Split the target description information to determine at least one first target text segment.

[0032] It can be understood that when splitting the target description information and correspondingly determining the first target text segments, one target description information may correspond to multiple first target text segments, that is, the file to be stored corresponds to multiple first target text segments. In this way, analysis and processing can be performed based on each first target text segment respectively to improve the accuracy of identifying the file to be stored.

[0033] In one embodiment, splitting the target description information to determine at least one first target text segment includes: Splitting the target description information according to a preset battery swapping operation and maintenance type to determine at least one initial text segment; Further splitting the multiple initial text segments according to the hierarchical information and / or the degree of context semantic association to determine at least one first target text segment.

[0034] Optionally, the preset battery swapping operation and maintenance types can include the operation and maintenance data types of the battery swapping station and the repair data types of the battery swapping station. In this way, based on different battery swapping operation and maintenance types, the corresponding splitting scales will also be different. For example, for the operation and maintenance data types of the battery swapping station, the splitting scale can be set to an interval of 500 - 800 characters; while for the repair data types of the battery swapping station, the splitting scale can be set to an interval of 500 - 1200 characters.

[0035] Among them, the hierarchical information reflects the organizational manner and the differences in importance degree of the text content; the degree of semantic association in context refers to the tightness of semantic association between text segments. In this way, during the splitting process, the obtained initial text segments can be further refined according to two metrics, namely the hierarchical information and the degree of semantic association in context in the initial text segments. At the same time, for at least one first target text segment obtained after splitting, an information of metadata can be correspondingly created to store some characteristic attributes about each first target text segment, including: related information such as the parent document of the source, the context document, the theme, the content keyword, the category label, etc.

[0036] Step S103: Determine at least one index vector of the file to be stored according to at least one first target text segment, so as to represent the file to be stored based on at least one index vector.

[0037] Optionally, determine the information expressed by the first target text segment and the relative information of the first target text segment corresponding to the entire file to be stored (that is, this segment is a certain part expressed by the entire article and the relationship with the entire article, etc.), and then index the first target text segment through a vectorization model to convert the first target text from a character form into a vector form in computer and mathematics, which can better understand and process natural language data, thereby improving the accuracy of information retrieval and question-answering systems.

[0038] In one embodiment, determining at least one index vector of the file to be stored according to at least one first target text segment includes: Determine the index vector corresponding to at least one first target text segment in the file to be stored; Determine the index vector corresponding to at least one first target text segment as at least one index vector of the file to be stored.

[0039] It can be understood that after splitting the file to be stored, one or more first target text segments corresponding to the file to be stored can be determined. Here, when determining the index vector corresponding to each first target text segment, the file to be stored can be traced and determined based on the index vector corresponding to each first target text segment. Therefore, a file to be stored can correspond to multiple index vectors.

[0040] In one embodiment, determining the index vector corresponding to at least one first target text segment in the file to be stored includes: Determine the third description information that is connected to the first target text segment; Determine the summary information of the first target text segment according to the third description information and the first target text segment; determining a query language of the first target text segment; An index vector of the first target text segment is determined according to the summary information and the query language.

[0041] Optionally, the context information inherited from the first target text segment is determined, and the third description information of the first target text segment is determined based on the context information, specifically considering the logical relationship and semantic association between these features, so as to construct the index vector more comprehensively and accurately. Among them, the context information inherited from the first target text segment can be other target text segments inherited from the first target text segment, or it can also be other data information associated with the first target text segment. For example, there may be a specific correspondence between "battery model" and "replacement tool", and different battery models may require different replacement tools. Optionally, after obtaining the third description information, the third description information and the data content in the first target text segment are integrated to determine it as summary information of the first target text segment.

[0042] Optionally, based on the first target text segment, a preset number of query languages are determined by corresponding outputs of the large model. The query language is generally expressed in the form of a question sentence. Thus, in subsequent matching, if a match with the query language is found, the first target text segment can be determined. The query language should be the most appropriate question sentence for the first target text segment.

[0043] In one embodiment, determining the query language of the first target text segment includes: A preset number of query languages related to the first target text segment are determined according to the first target text segment and the third description information.

[0044] Exemplarily, the split first target text segment and the third description information are input into the large model, and the large model is asked to output 5 questions for this text segment, and then the contents of these 5 questions are saved in the metadata of this segment.

[0045] In summary, in the data processing method for battery swapping operation and maintenance provided by the above embodiment, files are represented by index vectors, which can more conveniently generate different types of data outputs according to actual needs, while improving the accuracy of data output.

[0046] It can be understood that the multiple large models proposed in this application are based on the gateway interface as the intermediate layer, and are uniformly managed and scheduled accordingly. In this way, the upper-level application does not need to interact directly with each model, but calls it through the gateway. In this way, when changing the model, only the gateway part needs to be adjusted without changing other code data.

[0047] Among them, the large model can be a Large Language Model (LLM), which is an artificial intelligence-based model capable of understanding and generating natural language text. It is trained with a large amount of data to learn the structure, grammar, semantics, and context relationships of the language, and thus can perform tasks such as answering questions, writing articles, and translating languages. Common large models include the GPT series of OpenAI (such as GPT-3, GPT-4) and other similar models.

[0048] In one embodiment, the method further includes: Obtain the request problem for battery swapping operation and maintenance; According to a preset matching method, determine at least one index vector that matches the request problem; According to at least one index vector, determine the corresponding second target text segment; According to the second target text segment, determine the target file for obtaining battery swapping operation and maintenance in the database of battery swapping operation and maintenance.

[0049] As Figure 2 shown, after obtaining the problem, the obtained problem can be directly used as the request problem for battery swapping operation and maintenance, and the matching process of the data can be performed. Or, the obtained problem can also be rewritten to obtain a request problem for battery swapping operation and maintenance with a clearer and more accurate intention. Specifically, when rewriting the obtained problem, it includes: 1. According to the context of the conversation and the user's problem, directly output the rewritten problem through the large model to obtain the request problem for battery swapping operation and maintenance. Here, the rewritten content is mainly reference resolution and disambiguation. 2. Segment the user's problem, and then for some nouns, recall the proper nouns from the proper noun library of the battery swapping station by means of similarity matching to perform rewriting and determine the request problem for battery swapping operation and maintenance.

[0050] Optionally, after determining the request problem of the battery swap operation and maintenance, a recall operation is performed from the database of the battery swap operation and maintenance, such as the ElasticSearch database. Here, vector matching calculations can be performed through two methods: BM25 and vector matching. Among them, BM25 recall determines a preset number of index vectors that match in the database, such as obtaining the first 10 index vectors, and further determines the corresponding second target text segment based on the determined index vector. Vector matching controls the number and effect of recalled index vectors by controlling the threshold, and the threshold setting here is set through multiple debugging. Here, vector matching includes keyword matching, semantic understanding matching, and matching threshold matching determination. Specifically, in keyword-based matching, a matching value is calculated based on the number and weight of matched keywords to further determine the corresponding index vector; in semantic understanding-based matching, a numerical value representing the degree of matching is obtained through a semantic similarity algorithm to further determine the corresponding index vector. Set a matching threshold. When the calculated matching score is higher than the threshold, the corresponding index vector is determined as the index vector that matches the request problem.

[0051] In one embodiment, for the recalled second target text segments of dozens to hundreds, a rearrangement model can be used to rearrange these second target text segments. The rearrangement method is to evaluate the relevance of all texts to the semantic content of the question, and finally select the five second target text segments in the group as the final recalled text. Optionally, after determining the second target text segment, the corresponding target file can be directly determined in the battery swap operation and maintenance database based on the second target text segment.

[0052] In one embodiment, after obtaining the target file of battery swap operation and maintenance, the target file, the request question of battery swap operation and maintenance, and the context content related to the request question can be input into the big model, and the ability of the big model can be used to directly output the answer to the question, thereby achieving the effect of answering the user's question.

[0053] In one embodiment, the method further comprises: Obtain the target file as the answer result, and evaluate the accuracy of the answer result based on the preset test data set, including: Determine the accuracy of the answer based on the number of questions answered correctly and the total number of questions asked; and / or, In a multi-round dialogue, the accuracy of the answer result is determined based on the number of multi-round dialogues without errors and the total number of multi-round dialogues.

[0054] Optionally, for the knowledge questions of battery swapping operation and maintenance, the accuracy of the answer results can be evaluated based on a pre-set test data set. According to the number of correctly answered questions and the total number of questions, the accuracy of the answer results can be determined, which can be expressed by the formula: Accuracy = Number of correctly answered questions about battery swapping station operation and maintenance / Total number of questions and answers.

[0055] Optionally, in a multi-round conversation, as long as there is one wrong answer, it is determined that the answer in this conversation is incorrect. In this way, according to the number of multi-round conversations without errors and the total number of multi-round conversations, the accuracy of the answer results is determined, which is specifically expressed by the formula: Accuracy = Number of correct answers about battery swapping station operation and maintenance / Number of multi-round conversations.

[0056] Based on the same inventive concept as the foregoing embodiments, a data processing method for battery swapping operation and maintenance provided by the present application will be described in detail below through a specific example. As Figure 3 shown, through the natural language understanding of the large model and the knowledge capture ability of the context, the multi-modal interpretation ability is added, which improves the accuracy and relevance of the answer.

[0057] In order to answer questions related to battery swapping station operation and maintenance during the question-and-answer process, this knowledge belongs to domain-private knowledge and has the attribute of professional customization. Therefore, a data processing pipeline for processing the files to be stored is created to process the files to be stored specifically Figure 4 shown, for the user can upload various types of files to be stored, after passing through the data processing process, it is convenient for comprehensive knowledge retrieval and recall, and finally improves the accuracy of the entire question and answer.

[0058] A. First, the files to be stored will be classified and labeled according to classification labels such as battery swapping station operation and maintenance, repair, maintenance, and battery swapping process for all types, and stored in different directories according to the classification.

[0059] B. Process the content in the files to be stored. Identify all the text content in the files by loading the text and by OCR recognition, and then process all the text content through the large model, and the output result is in the general format of markdown.

[0060] C. For the picture content in the files to be stored, identify the pictures through the program and automatically save the picture sources in the database of battery swapping operation and maintenance. At the same time, in the context of the processed text content, embed the picture format in markdown format and the corresponding picture name, and add the information text for picture explanation. These information texts are summarized from the context where the pictures are located and the information in the pictures.

[0061] D. Split the text according to the text content in markdown format to determine at least one initial text segment. The splitting scale is distinguished according to the type of text. For example, for the document on problem-solving in the operation and maintenance process of battery swapping stations, it is recommended that the splitting scale be in the range of 500-800 characters; for the operation document of battery swapping station maintenance, it is recommended that the splitting scale be in the range of 500-1200 characters.

[0062] E. During the splitting process, the document can be further refined and split according to two metrics: the hierarchy of the document and the degree of semantic association of the context, that is, determine the first target text segment. At the same time, for the split document, create a metadata information corresponding to it, which stores some characteristic attributes of the first target text segment, including: the parent document of the source, the context document, the theme, the content keyword, the category label and other relevant information.

[0063] F. In order to improve the accuracy and recall rate of the subsequent retrieved data, the data processing can be further optimized. The custom operations are as follows: 1. For the split first target text segment, use the large model to identify the text segment and the context of the text segment, and let the model output 5 query languages for the first target text segment, and then save the content of these 5 query languages in the metadata of this segment.

[0064] 2. For the split first target text segment, use the large model to identify the first target text segment and the context of the first target text segment, and let the model output a summary for the first target text segment. The summary information includes the content described in this text and the inheritance relationship with the context, and then save this summary information in the metadata of this segment.

[0065] 3. For the processed document content, use the encoding model based on the transformer architecture to vectorize the document to determine the corresponding index vector. Here, the encoding model is trained on the knowledge materials in the field of battery swapping stations, and the vectorized expression of the proper nouns in the battery swapping stations is more accurate. After vectorization, the first target text segment is stored in the ElasticSearch database for convenient use in subsequent data retrieval.

[0066] Continue to refer to Figure 3 , this embodiment configures the interface interaction of multiple models. Different models provide different capabilities, making the entire framework have diverse capabilities and strong scalability.

[0067] A. For the user's question, the user's question can be rewritten to determine the request question for battery swapping operation and maintenance. There are two solutions for the rewriting method: 1. Based on the context of the conversation and the user's question, directly output the rewritten question through the large model to obtain the request question for battery swapping operation and maintenance. Here, the rewritten content mainly refers to anaphora resolution and disambiguation.

[0068] 2. By tokenizing the user's question, and then for some nouns, recall the proper nouns from the battery swapping station proper noun library through similarity matching to rewrite and determine the request question for battery swapping operation and maintenance.

[0069] B. Perform a recall operation on the rewritten question from the ElasticSearch database. Here, the recall methods adopt two schemes: BM25 and vector matching. BM25 recall can control the recall of the top 10 data, and the vector matching scheme controls the recall quantity and effect by controlling the threshold. The threshold setting here is determined through multiple debuggings.

[0070] C. For the dozens to hundreds of recalled data, use a re-rank model to re-rank these texts. The re-rank method is evaluated according to the semantic content fit of all texts and the question sentence. Finally, take 5 texts in the group as the final recalled texts.

[0071] D. Then output the recalled texts, question sentences, conversation context, etc. to the large model, and use the capabilities of the large model to directly output the answers to the questions, achieving the effect of answering the user's questions.

[0072] In one implementation, this embodiment uses a pre-trained model as the base model to adapt to various large models. However, in order to improve the effectiveness of this system in the field of battery swapping station operation and maintenance, a dedicated model for the vertical domain can be correspondingly set. Use the Q&A data in the dedicated domain as training data to perform secondary training on the general model. Through a certain number of trainings, use the prepared evaluation data to evaluate the model, evaluate the learning quality and overall capabilities of the model, and use the obtained model here. Specifically, it includes: A. For the collected documents related to battery swapping stations, these documents contain various relevant data such as the operation and maintenance, maintenance, operation, and business of battery swapping stations, and perform data processing in the form of question-answer. The processing method is set according to the actual text: 1. If the text content is directly a pair of question and answer, split it directly according to the label.

[0073] 2. For narrative documents, input them into the large model and let the large model generate a certain number of questions and answers according to the content.

[0074] B. According to the prepared certain number of question and answer pair data, perform fine-tuning training on the selected model. After a certain number of trainings, evaluate the accuracy of the model.

[0075] In summary, in the data processing method for battery swapping operation and maintenance provided by the above embodiments, representing files through index vectors can more conveniently generate different types of data outputs according to actual requirements, and at the same time improve the accuracy of data outputs.

[0076] Based on the same inventive concept as the foregoing embodiments, an embodiment of the present invention provides a computing device, as Figure 5 shown. The computing device includes: a processor 210 and a memory 211 storing a computer program; wherein, Figure 5 the processor 210 shown in does not refer to the number of processors 210 being one, but only refers to the positional relationship of the processor 210 relative to other components. In actual applications, the number of processors 210 can be one or more; similarly, Figure 5 the memory 211 shown in has the same meaning, that is, it only refers to the positional relationship of the memory 211 relative to other components. In actual applications, the number of memories 411 can be one or more. When the processor 210 runs the computer program, the above data processing method for battery swapping operation and maintenance is implemented.

[0077] The computing device may further include: at least one network interface 212. Each component in the computing device is coupled together through a bus system 213. It can be understood that the bus system 213 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 213 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clear illustration, in Figure 5 all kinds of buses are labeled as the bus system 213.

[0078] Among them, the memory 211 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memory 211 described in the embodiments of the present invention is intended to include but not limited to these and any other suitable types of memories.

[0079] The memory 211 in the embodiments of the present invention is used to store various types of data to support the operation of the computing device. Examples of such data include: any computer programs for operating on the computing device, such as operating systems and application programs; contact data; phone book data; messages; pictures; videos, etc. Among them, the operating system contains various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application programs can include various application programs, such as a media player, a browser, etc., for implementing various application services. Here, the program for implementing the method of the embodiments of the present invention can be included in the application programs.

[0080] Based on the same inventive concept as the foregoing embodiments, this embodiment also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. The computer-readable storage medium can be a ferromagnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, etc. When the computer program stored in the computer-readable storage medium is run by a processor, it implements the data processing method for battery swapping operation and maintenance applied to the above computing device. For the specific step flow implemented when the computer program is executed by the processor, please refer to Figure 1 the description of the illustrated embodiments, which will not be repeated here.

[0081] The technical features of the above-described embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0082] In this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion. In addition to the elements listed, it may also include other elements not expressly listed.

[0083] As described above, this is only a specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A data processing method for battery swapping operation and maintenance, characterized in that, include: Obtain the files to be stored for battery swap operation and maintenance, and determine the target description information of the files to be stored; Splitting the target description information to determine at least one first target text segment; At least one index vector of the file to be stored is determined according to the at least one first target text segment, so as to represent the file to be stored based on the at least one index vector.

2. The method according to claim 1, characterized in that, The file to be stored includes image data, and the obtaining of the file to be stored for battery swapping operation and maintenance and determining target description information of the file to be stored include: Determining first description information carried by the image data in text form; Acquire second description information of the context associated with the image data in the form of text; The first description information and the second description information are converted into a target format and determined as target description information of the file to be stored.

3. The method according to claim 1, wherein The step of splitting the target description information to determine at least one first target text segment includes: According to the preset battery swap operation and maintenance type, the target description information is split and processed to determine at least one initial text segment; The multiple initial text segments are further split according to the hierarchical information and / or the degree of association of contextual semantics to determine at least one first target text segment.

4. The method according to claim 1, wherein The step of determining at least one index vector of the file to be stored based on the at least one first target text segment includes: Determining an index vector corresponding to at least one first target text segment in the file to be stored; The index vector corresponding to the at least one first target text segment is determined as at least one index vector of the file to be stored.

5. The method according to claim 4, wherein The determining of the index vector corresponding to the at least one first target text segment in the file to be stored includes: Determining third description information that continues with the first target text segment; determining summary information of the first target text segment according to the third description information and the first target text segment; determining a query language of the first target text segment; An index vector of the first target text segment is determined according to the summary information and the query language.

6. The method according to claim 5, characterized in that, The determining of the query language of the first target text segment includes: A preset number of query languages related to the first target text segment are determined according to the first target text segment and the third description information.

7. The method according to claim 1, wherein The method further comprises: Obtaining requests for battery swap operations and maintenance; Determining at least one index vector that matches the request question according to a preset matching method; Determining a corresponding second target text segment according to the at least one index vector; According to the second target text segment, the target file for battery swapping operation and maintenance is determined and obtained in the battery swapping operation and maintenance database.

8. The method according to claim 7, wherein The method further comprises: Obtain the target file as the answer result, and evaluate the accuracy of the answer result based on the preset test data set, including: Determining the accuracy of the answer results based on the number of questions answered correctly and the total number of questions; and / or, In a multi-round dialogue, the accuracy of the answer result is determined based on the number of multi-round dialogues without errors and the total number of multi-round dialogues.

9. A computing device, characterized in that, include: A processor and a memory for storing executable instructions; wherein the processor is configured to execute the instructions to implement the data processing method for battery swapping operation and maintenance as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor, the data processing method for battery swapping operation and maintenance as described in any one of claims 1-8 is implemented.

Images