Data retrieval method and device based on index platform, electronic equipment and medium

Abstract

The invention provides a data retrieval method and device based on an index platform, electronic equipment and a medium, and the method comprises the steps: obtaining a retrieval problem input by a user, and extracting a keyword of the retrieval problem based on a pre-trained first large language model; wherein the keyword comprises time information and at least one piece of index information; performing vectorization processing on the keyword based on the vector model after fine tuning to obtain a vector of the keyword; performing retrieval and calculation on a pre-constructed index platform based on the keyword and the vector of the keyword to obtain a retrieval result of the index information; wherein the retrieval result of the index information comprises calculation results of single production data or multiple pieces of production data; the index platform is connected with different data sources to obtain production data, and various computational logics are built in the index platform and used for calculating the production data. According to the method, the accuracy of data retrieval is improved, and meanwhile the utilization rate of production data is increased.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data retrieval method, device, electronic equipment and medium based on an indicator platform. Background Art

[0002] With the rapid development of large language model technology, its application has gradually penetrated into all walks of life, and the industrial and mining fields are no exception. Large language models are gradually penetrating into the production process and playing an increasingly important role.

[0003] Nowadays, intelligent devices such as sensors are widely used in the production process of industrial and mining enterprises, and these devices continuously generate massive amounts of production data. The industrial and mining fields mainly rely on these basic production data to drive business decisions, and store the data in the database through ETL (extraction, transformation, loading) or ELT (extraction, loading, transformation) technology. In the current industrial production field, although data collection and storage technology is quite mature, how to efficiently use these massive production data is still an urgent problem to be solved. In particular, basic production data and data that have been preliminarily processed are often not fully utilized and are not closely associated with actual production conditions, such as calculating unit consumption or providing month-on-month analysis. In addition, the current system lacks an effective mechanism to parse retrieval requests and perform complex data queries and indicator summaries based on this, so there may be situations where the retrieval results do not match the retrieval requests. Summary of the invention

[0004] In view of this, the object of the present invention is to provide a data retrieval method, device, electronic device and medium based on an indicator platform, so as to improve the accuracy of data retrieval and improve the utilization rate of production data.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a data retrieval method based on an indicator platform, comprising: obtaining a retrieval question input by a user, and extracting keywords of the retrieval question based on a pre-trained first language model; wherein the keywords include: time information and at least one indicator information; vectorizing the keywords based on a fine-tuned vector model to obtain a vector of the keyword; searching and calculating based on the keywords and the vectors of the keywords on a pre-built indicator platform to obtain retrieval results of the indicator information; wherein the retrieval results of the indicator information include: calculation results of a single production data or multiple production data; the indicator platform is connected to different data sources to obtain production data, and the indicator platform has multiple built-in calculation logics, and the calculation logic is used to calculate the production data.

[0006] Optionally, after extracting the keywords of the retrieval question based on the pre-trained first language model, it also includes: comparing the indicator information with a preset synonym library, and converting the indicator information into a corresponding indicator alias or indicator description.

[0007] Optionally, the indicator platform includes a retrieval library, which includes: the indicator name, indicator description and indicator alias of each indicator, and the vectors corresponding to the indicator name, indicator description and indicator alias; searching and calculating based on keywords and keyword vectors on a pre-built indicator platform to obtain retrieval results for each indicator information, including: for each indicator information, calculating the similarity between the indicator information and the indicator name, indicator description and indicator alias based on the retrieval library, and calculating the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias to obtain multiple indicator recall results; sorting the indicator recall results based on the similarity, and determining the target retrieval indicator corresponding to each indicator information based on the sorting result; and obtaining the retrieval results from the indicator platform based on the target retrieval indicator.

[0008] Optionally, retrieval results are obtained from the indicator platform based on the target retrieval indicator, including: obtaining production data associated with the target retrieval indicator from the indicator platform as the retrieval result of the indicator information; or, obtaining production data associated with the target retrieval indicator from the indicator platform, and performing calculations based on the production data and calculation logic to obtain the retrieval result of the indicator information.

[0009] Optionally, it also includes: obtaining historical conversation records of the first largest language model; inputting the historical conversation records, retrieval questions and indicator recall results into the first largest language model to sort the indicator recall results, and determining the first indicator recall result in the sorted results as the target retrieval indicator.

[0010] Optionally, retrieval and calculation are performed on a pre-built indicator platform based on keywords and keyword vectors to obtain retrieval results for each indicator information, including: for each indicator information, based on the retrieval library, calculating the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculating the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; determining the indicator recall result based on the similarity, a preset similarity threshold and a maximum recall number, and inputting the indicator recall result and the retrieval question into a predetermined rearrangement model, returning the first indicator recall result after rearrangement as the target retrieval indicator; and obtaining the retrieval result from the indicator platform based on the target retrieval indicator.

[0011] Optionally, retrieval and calculation are performed on a pre-built indicator platform based on keywords and keyword vectors to obtain retrieval results for each indicator information, including: for each indicator information, based on the retrieval library, calculating the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculating the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; determining the indicator recall result based on the similarity, a preset similarity threshold and a maximum recall number; inputting the indicator recall result and the retrieval question into the second largest language model to obtain the target retrieval indicator; and obtaining the retrieval result from the indicator platform based on the target retrieval indicator.

[0012] In a second aspect, the present invention provides a data retrieval device based on an indicator platform, comprising: an indicator extraction module, used to obtain a retrieval question input by a user, and extract keywords of the retrieval question based on a pre-trained first language model; wherein the keywords include: time information and at least one indicator information; a vectorization module, used to vectorize the keywords based on a fine-tuned vector model to obtain a vector of the keyword; a retrieval module, used to retrieve and calculate based on the keywords and the vectors of the keywords on a pre-built indicator platform to obtain retrieval results of the indicator information; wherein the retrieval results of the indicator information include: calculation results of a single production data or multiple production data; the indicator platform is connected to different data sources to obtain production data, and the indicator platform has multiple built-in calculation logics, and the calculation logic is used to calculate the production data.

[0013] In a third aspect, the present invention provides an electronic device comprising a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of any one of the methods provided in the first aspect above.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods provided in the first aspect are executed.

[0015] The present invention brings the following beneficial effects: The data retrieval method, device, electronic device and medium based on the index platform provided by the present invention first obtain the search question input by the user, and extract the keywords of the search question based on the pre-trained first language model; wherein the keywords include: time information and at least one index information; then the keywords are vectorized based on the fine-tuned vector model to obtain the vector of the keywords; finally, the search and calculation are performed on the pre-built index platform based on the keywords and the vector of the keywords to obtain the search results of the index information; wherein the search results of the index information include: the calculation results of a single production data or multiple production data. In the above method, the search question input by the user is first analyzed and keywords are extracted using the pre-trained first language model, and then the index platform is used to perform index retrieval based on the keywords, thereby improving the accuracy of data retrieval; at the same time, the index platform is used for retrieval in the above method, and the index platform can be connected with different data sources to obtain production data, and has built-in multiple calculation logics, which can calculate the production data to obtain more comprehensive indicators, thereby integrating and deeply utilizing the production data and improving the utilization rate of the production data.

[0016] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the preferred embodiments are specifically mentioned below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A flow chart of a data retrieval method based on an indicator platform provided by an embodiment of the present invention; Figure 2 A flowchart of an indicator platform query interaction provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a data retrieval device based on an indicator platform provided by an embodiment of the present invention; Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. 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.

[0021] At present, in the current industrial production field, although data collection and storage technology is quite mature, how to efficiently use these massive production data is still an urgent problem to be solved. In particular, basic production data and data that have been preliminarily processed are often not fully utilized and cannot be closely linked to actual production conditions, such as calculating unit consumption or providing month-on-month analysis. In addition, the current system lacks an effective mechanism to parse retrieval requests and perform complex data queries and indicator summaries based on this, so there may be situations where the retrieval results do not match the retrieval requests.

[0022] Based on this, the embodiments of the present invention provide a data retrieval method, device, electronic device and medium based on an indicator platform, which can improve the accuracy of data retrieval and improve the utilization rate of production data.

[0023] To facilitate understanding of this embodiment, the indicator platform disclosed in the embodiment of the present invention is first introduced.

[0024] For the large amount of production data generated during the industrial and mining production process, the existing data center is first used for data processing, and the data is stored in the database through the ETL or ELT process. The unified interface provided by the data center enables the indicator platform to directly access real-time and historical data services, thereby ensuring the consistency and convenience of data queries. At the same time, combined with resource conditions, the indicator platform can also have built-in multiple calculation logics for calculating production data. Specifically, programming logic can be applied to perform data calculations for specific scenarios, such as unit consumption and output statistics. The indicator platform provided in the embodiment of the present invention has the following functions: (1) Universal docking: The indicator platform provides flexible docking functions or operators to support connections with different data sources (including data middle platforms and databases such as Elasticsearch, MySQL, and MongoDB).

[0025] (2) Dynamic query processing: By writing scripts or functions (i.e., calculation logic), users can create query result processing functions or operators that can be dynamically combined to meet different data analysis requirements. That is, users can combine processing functions or operators according to actual needs to obtain the calculation logic of the required indicator information. For example, if users need to calculate the unit consumption within a period of time, they can select the processing function of total consumption divided by time, and the total consumption can be obtained by summing the consumption within the period of time.

[0026] (3) Indicator formatting: To ensure that the query results are presented in a manner suitable for subsequent arrangement and combination, the standard format of each indicator can be configured in the indicator platform. After obtaining the production data corresponding to the indicator, calculations and data conversion can be performed on it so that it can be output to the user in a standard format. For non-real-time statistical data, the indicator platform supports periodic calculations, real-time calculations, and data conversion operations.

[0027] (4) Data association: The indicator platform is not limited to the data in the data center, but can also configure related data items according to needs to help users understand the causes of abnormal data and provide guidance in combination with the on-site knowledge graph. Specifically, in actual production, the production data of a certain device may be affected by the data of other devices. The indicator platform can associate them and feed back the related indicators to the user when querying the data, so that the user can perform abnormal analysis.

[0028] (5) Public indicator management: Users can define and manage public indicators of concern to the factory through the indicator platform, and flexibly use these indicators and their combinations according to changes in production needs.

[0029] (6) Query permission control: Users can use the indicator platform to set permissions for which indicators can be queried through natural language, thereby enhancing data security and privacy protection.

[0030] (7) Multi-type data storage: Supports saving query results to various storage tools such as time series databases, relational databases, or non-relational databases.

[0031] Furthermore, the indicator platform includes a retrieval library, which includes: the indicator name, indicator description and indicator alias of each indicator, and the vectors corresponding to the indicator name, indicator description and indicator alias. In the specific implementation, when configuring the indicator name, it is necessary to ensure that all indicator name settings comply with industry standards. At the same time, an indicator alias and indicator description can be added to each indicator name to facilitate more efficient and accurate retrieval of indicators. The indicator name, indicator description, and indicator alias on the indicator platform are quantized using a fine-tuned vector model, and this information (including the indicator name, indicator description, and indicator alias before and after quantization) is stored in a database that supports both text retrieval and vector retrieval to obtain a retrieval library (Elasticsearch can be selected in the embodiment of the present invention, but is not limited to this database).

[0032] When configuring the search library, in order to improve the search efficiency, you can specify the searchable fields (i.e. the fields corresponding to the indicator name, indicator description, and indicator alias). For example: if the accurate data can be retrieved by the indicator name and indicator alias, there is no need to search by the indicator description. You can set the indicator description and its corresponding vector field as non-searchable. However, the search function of the indicator name cannot be ignored, that is, the indicator name and its corresponding vector field cannot be set as non-searchable. After the construction of the indicator platform is completed, the indicator platform will have powerful natural language query capabilities, which can help users efficiently retrieve the required information.

[0033] Next, a data retrieval method based on an indicator platform disclosed in an embodiment of the present invention is introduced in detail. The method can be executed by an electronic device, such as a smart phone, a computer, a tablet computer, etc. Figure 1 The flowchart of a data retrieval method based on an indicator platform is shown, which illustrates that the method mainly includes the following steps S101 to S103: Step S101: Obtain a search question input by a user, and extract keywords of the search question based on a pre-trained first language model.

[0034] In one embodiment, after the user inputs a search question (natural language question), the search question is first parsed using a pre-trained first language model with at least 7B level parameters to extract keywords such as time information and indicator information. In this embodiment, the indicator information can be one or more, depending on the content of the query question. For example, if the query question input by the user is "Please help me find the total output in the past week", the time information that can be extracted by the pre-trained first language model is "past week" and the indicator information is "total output".

[0035] Step S102: vectorizing the keyword based on the fine-tuned vector model to obtain a vector of the keyword.

[0036] In one embodiment, the fine-tuned vector model uses the Chinese vector library as a basis and is fine-tuned for the usage habits of a specific industry to create a vector model that can understand industry terms. Daily industry data and questions, including the industry terms involved, are collected to train and optimize the vector model. Based on this, the fine-tuned vector model can be used to vectorize the extracted keywords to obtain keyword vectors.

[0037] Step S103: Search and calculate based on the keywords and keyword vectors on a pre-built indicator platform to obtain a search result of the indicator information.

[0038] Among them, the retrieval results of the indicator information include: the calculation results of a single production data or multiple production data. In one embodiment, when performing data retrieval, the retrieval can be performed simultaneously according to the text content of the keyword and the vector of the keyword, the similarity between the keyword and each indicator in the indicator platform is calculated, and the retrieval results are sorted according to the similarity. For the query of a single indicator, the first item after sorting is obtained by default as the retrieved indicator, and then the production data corresponding to the retrieved indicator is obtained. If the indicator that the user needs to retrieve is the existing production data, the production data corresponding to the retrieved indicator is directly output to the user as the retrieval result; if the indicator that the user needs to retrieve needs to be calculated, the associated production data is obtained for calculation to obtain the final retrieval result. For example: if the user needs to query yesterday's output, yesterday's output is directly obtained and output to the user; if the user needs to query the total output of the past week, the output of each day in the past week is obtained and summed to obtain the total output, and the total output is output to the user.

[0039] The data retrieval method based on the indicator platform provided by the present invention first uses the pre-trained first language model to analyze the search question input by the user and extract keywords, and then uses the indicator platform to perform indicator retrieval based on the keywords, thereby improving the accuracy of data retrieval; at the same time, the indicator platform is used for retrieval in the above method, and the indicator platform can be connected to different data sources to obtain production data, and has built-in multiple calculation logics, which can calculate the production data to obtain more comprehensive indicators, thereby integrating and deeply utilizing the production data and improving the utilization rate of the production data.

[0040] In one embodiment, in order to improve the accuracy of data retrieval, after extracting the keywords of the retrieval question based on the pre-trained first language model, the embodiment of the present invention also includes: comparing the indicator information with a preset synonym library, and converting the indicator information into a corresponding indicator alias or indicator description.

[0041] In specific implementation, a synonym library of indicators is pre-set in the indicator platform, and the synonym library includes commonly used aliases and descriptions corresponding to the indicator names. After extracting the keywords of the search questions input by the user, the keywords of the extracted search questions can be compared with the synonym library. For some local or industry-specific descriptions, they are converted into commonly used description methods, that is, converted into commonly used indicator aliases or indicator descriptions to ensure correct retrieval.

[0042] In one embodiment, for the aforementioned step S103, that is, when searching and calculating based on the keywords and the keyword vectors on the pre-built indicator platform to obtain the search results of each indicator information, the following three methods may be used, including but not limited to: Method 1: First, for each indicator information, based on the retrieval library, calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculate the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias, and obtain multiple indicator recall results; then sort the indicator recall results based on the similarity, and determine the target retrieval indicator corresponding to each indicator information based on the sorting results; finally, obtain the retrieval results from the indicator platform based on the target retrieval indicator.

[0043] In the specific implementation, first calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias in the retrieval library, and the similarity between the indicator information vector and the vector corresponding to the indicator name, indicator description and indicator alias in the retrieval library, and obtain multiple indicator recall results (i.e. indicators similar to the indicator information) according to the similarity calculation results; then sort the indicator recall results in descending order of similarity, and use the first indicator recall result after sorting as the target retrieval indicator (i.e. the indicator that the user wants to query); finally, obtain the production data associated with the target retrieval indicator from the indicator platform as the retrieval result. If the retrieval question includes multiple indicator information, perform the above operations for each indicator information to obtain the corresponding retrieval result.

[0044] Specifically, when obtaining search results from the indicator platform based on the target search indicator, there are two situations: (1) Obtain production data associated with the target retrieval indicator from the indicator platform as the retrieval result of the indicator information.

[0045] In specific implementation, if the indicator that the user needs to retrieve is existing production data, the production data corresponding to the retrieved indicator is directly output to the user as a search result.

[0046] (2) Obtain the production data associated with the target retrieval indicator from the indicator platform, and perform calculations based on the production data and calculation logic to obtain the retrieval results of the indicator information.

[0047] In specific implementation, if the index that the user needs to retrieve needs to be calculated, the associated production data is obtained, and the final search result is obtained by calculating according to the preset calculation logic.

[0048] In order to further improve the accuracy of retrieval, the above method also includes: obtaining historical conversation records of the first largest language model; inputting the historical conversation records, retrieval questions and indicator recall results into the first largest language model to sort the indicator recall results, and determining the first indicator recall result in the sorted results as the target retrieval indicator.

[0049] In specific implementation, after obtaining the indicator recall result, context retrieval can be performed to obtain the historical conversation record of the first language model. In this embodiment, 5 to 10 rounds of historical conversation records can be referred to; then the indicator recall result is analyzed and sorted in combination with the historical conversation record and the search question, and the first indicator recall result after sorting is determined as the target search indicator, so as to more accurately infer the specific indicator referred to by the current search question. In the embodiment of the present invention, combined with historical data analysis, the effect of online real-time analysis can be enhanced to provide users with more intelligent services.

[0050] Method 2: First, for each indicator information, based on the retrieval library, calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculate the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; then, determine the indicator recall result based on the similarity, the preset similarity threshold and the maximum recall number, and input the indicator recall result and the retrieval question into a predetermined rearrangement model, and return the first indicator recall result after rearrangement as the target retrieval indicator; finally, obtain the retrieval result from the indicator platform based on the target retrieval indicator.

[0051] In the specific implementation, first calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias in the search library, as well as the similarity between the indicator information vector and the vector corresponding to the indicator name, indicator description and indicator alias in the search library; then, according to the pre-set similarity threshold, recall the indicators with similarity greater than the similarity threshold to obtain the indicator recall result; at the same time, set the maximum recall number. If the indicators with similarity greater than the similarity threshold exceed the maximum recall number, select the indicators with the maximum recall number as the indicator recall result. For example: assuming the similarity threshold is 0.5 and the maximum recall number is 10, if the number of indicators with similarity greater than 0.5 is less than 10, all indicators with similarity greater than 0.5 will be recalled; if the number of indicators with similarity greater than 0.5 is greater than 10, 10 indicators will be recalled from the largest to the smallest similarity as the indicator recall result.

[0052] Furthermore, the indicator recall results and the retrieval question are input into a predetermined re-ranking model (such as a fine-tuned rereanker re-ranking model), the indicator recall results are re-ranked, and the first indicator recall result after re-ranking is returned as the target retrieval indicator; finally, the retrieval result of the target retrieval indicator is obtained from the indicator platform.

[0053] Method 3: First, for each indicator information, based on the retrieval library, calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculate the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; then, determine the indicator recall result based on the similarity, the preset similarity threshold and the maximum recall number; input the indicator recall result and the retrieval question into the second largest language model to obtain the target retrieval indicator; finally, obtain the retrieval result from the indicator platform based on the target retrieval indicator.

[0054] In the specific implementation, first calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias in the retrieval library, and the similarity between the indicator information vector and the vector corresponding to the indicator name, indicator description and indicator alias in the retrieval library; then, according to the pre-set similarity threshold, recall the indicators with similarity greater than the similarity threshold to obtain the indicator recall result; at the same time, set the maximum recall number. If the indicators with similarity greater than the similarity threshold exceed the maximum recall number, the indicators with the maximum recall number are selected as the indicator recall result. Then, input the indicator recall result and the retrieval question into the second largest language model (a model with a higher order than the first largest language model), and select the indicator recall result that best meets the retrieval question as the target retrieval indicator through the second largest language model; finally, obtain the retrieval result of the target retrieval indicator from the indicator platform.

[0055] In addition, in the implementation of the present invention, indicators can be recalled based on ES segmentation and using retrieval auxiliary field data by setting common query words for indicators, for example: setting the common query word "cyclone" for heavy medium cyclone, and then using the common query words to recall indicators.

[0056] The above method provided by the embodiment of the present invention can realize flexible query and in-depth analysis of industrial and mining production data, not only can respond to specific questions about production data, but also provide contextual information of related data and provide necessary analysis and reasoning support.

[0057] For ease of understanding, the embodiment of the present invention also provides a specific indicator platform query interaction process, see Figure 2 As shown, first, the production equipment data of the production system and the production data reported by the user are obtained through the data middle platform, and the obtained data are processed; then an indicator platform is built based on the data middle platform, and indicator information is extracted from the indicator platform to create a retrieval library.

[0058] When the user enters a query question, the query question is quantified using the fine-tuned vector model, and the similarity of the indicator information in the query question is calculated based on the quantified query question and the retrieval library; then the indicator recall result is obtained by relevance sorting based on the similarity, and the indicator recall result with the top 1 similarity sorting is selected as the target retrieval indicator; or 10 rounds of historical conversation records are obtained from the question and answer history, and the indicator recall results based on similarity and historical conversation records are sorted through a large language model, and the top 1 indicator recall result is selected as the target retrieval indicator; finally, the corresponding production data is obtained from the indicator platform, and the retrieval results are analyzed using a data analysis tool set, including but not limited to: trend prediction, anomaly detection, etc., to help users better understand the meaning of the data and make more informed decisions based on it.

[0059] The above method provided by the embodiment of the present invention, combined with the powerful capabilities of the large language model, can conduct in-depth analysis and reasoning on production data, and provide users with more comprehensive data support; by integrating real-time data streams and historical records, it ensures the real-time and comprehensiveness of data analysis, and supports more accurate trend judgment and early warning mechanisms; through natural language processing technology, users do not need to have professional skills to operate, which greatly reduces the usage threshold and improves work efficiency; through refined management and optimization suggestions for various key indicators in the production process, it promotes the continuity and stability of production, and ultimately achieves the purpose of improving overall production efficiency.

[0060] For the data retrieval method based on the indicator platform provided in the above embodiment, the embodiment of the present invention also provides a data retrieval device based on the indicator platform, see Figure 3 The schematic diagram of the structure of a data retrieval device based on an indicator platform is shown, indicating that the device mainly includes the following parts: The indicator extraction module 301 is used to obtain the search question input by the user and extract the keywords of the search question based on the pre-trained first language model; wherein the keywords include: time information and at least one indicator information; A vectorization module 302, configured to perform vectorization processing on the keyword based on the fine-tuned vector model to obtain a vector of the keyword; The retrieval module 303 is used to perform retrieval and calculation on a pre-built indicator platform based on keywords and keyword vectors to obtain retrieval results of indicator information; wherein the retrieval results of indicator information include: calculation results of a single production data or multiple production data; the indicator platform is connected to different data sources to obtain production data, and the indicator platform has multiple built-in calculation logics, which are used to calculate the production data.

[0061] The data retrieval device based on the indicator platform provided by the present invention first uses the pre-trained first language model to analyze the search question input by the user and extract keywords, and then uses the indicator platform to perform indicator retrieval based on the keywords, thereby improving the accuracy of data retrieval; at the same time, the indicator platform is used for retrieval in the above-mentioned device, and the indicator platform can be connected to different data sources to obtain production data, and has built-in multiple calculation logics, which can calculate the production data to obtain more comprehensive indicators, thereby integrating and deeply utilizing the production data and improving the utilization rate of the production data.

[0062] In one embodiment, the above-mentioned device further includes: a comparison module, which is used to compare the indicator information with a preset synonym library and convert the indicator information into a corresponding indicator alias or indicator description.

[0063] In one embodiment, the indicator platform includes a retrieval library, which includes: the indicator name, indicator description and indicator alias of each indicator, and the vectors corresponding to the indicator name, indicator description and indicator alias; the above-mentioned retrieval module 303 is specifically used to: for each indicator information, based on the retrieval library, calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculate the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias, to obtain multiple indicator recall results; sort the indicator recall results based on the similarity, and determine the target retrieval indicator corresponding to each indicator information based on the sorting result; obtain the retrieval result from the indicator platform based on the target retrieval indicator.

[0064] In one embodiment, the above-mentioned retrieval module 303 is specifically used to: obtain production data associated with the target retrieval indicator from the indicator platform as the retrieval result of the indicator information; or, obtain production data associated with the target retrieval indicator from the indicator platform, and perform calculations based on the production data and calculation logic to obtain the retrieval result of the indicator information.

[0065] In one embodiment, the retrieval module 303 is specifically used to: obtain historical conversation records of the first largest language model; input the historical conversation records, retrieval questions and indicator recall results into the first largest language model to sort the indicator recall results, and determine the first indicator recall result in the sorted results as the target retrieval indicator.

[0066] In one embodiment, the above-mentioned retrieval module 303 is specifically used for: for each indicator information, based on the retrieval library, calculating the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculating the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; determining the indicator recall result based on the similarity, a preset similarity threshold and a maximum recall quantity, and inputting the indicator recall result and the retrieval question into a predetermined rearrangement model, returning the first indicator recall result after rearrangement as the target retrieval indicator; obtaining the retrieval result from the indicator platform based on the target retrieval indicator.

[0067] In one embodiment, the above-mentioned retrieval module 303 is specifically used to: for each indicator information, based on the retrieval library, calculate the similarity between the indicator information and the indicator name, indicator description and indicator alias, and calculate the similarity between the vector of the indicator information and the vector corresponding to the indicator name, indicator description and indicator alias; determine the indicator recall result based on the similarity, a preset similarity threshold and a maximum recall number; input the indicator recall result and the retrieval question into the second largest language model to obtain the target retrieval indicator; obtain the retrieval result from the indicator platform based on the target retrieval indicator.

[0068] It should be noted that the implementation principle and technical effects of the device provided in the embodiment of the present invention are the same as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0069] An embodiment of the present invention further provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above implementation methods.

[0070] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43, wherein the processor 40, the communication interface 43 and the memory 41 are connected via the bus 42; the processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0071] The memory 41 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0072] The bus 42 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0073] Among them, the memory 41 is used to store programs, and the processor 40 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0074] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 40. The above processor 40 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present invention can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and completes the steps of the above method in combination with its hardware.

[0075] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0076] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. 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., which can store program codes.

[0077] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A data retrieval method based on an indicator platform, characterized in that: include: Obtaining a search question input by a user, and extracting keywords of the search question based on a pre-trained first language model; wherein the keywords include: time information and at least one indicator information; Performing vectorization processing on the keyword based on the fine-tuned vector model to obtain a vector of the keyword; Based on the keywords and the vectors of the keywords, retrieval and calculation are performed on a pre-built indicator platform to obtain retrieval results of indicator information; wherein the retrieval results of the indicator information include: calculation results of a single production data or multiple production data; the indicator platform is connected to different data sources to obtain production data, and the indicator platform has multiple built-in calculation logics, and the calculation logic is used to calculate the production data.

2. The method according to claim 1, characterized in that After extracting the keywords of the search question based on the pre-trained first language model, the method further includes: The indicator information is compared with a preset synonym library, and the indicator information is converted into a corresponding indicator alias or indicator description.

3. The method according to claim 1, characterized in that The indicator platform includes a search library, which includes: an indicator name, an indicator description and an indicator alias of each indicator, and vectors corresponding to the indicator name, the indicator description and the indicator alias; Based on the keyword and the keyword vector, a search and calculation is performed on a pre-built indicator platform to obtain a search result for each indicator information, including: For each indicator information, based on the search library, the similarity between the indicator information and the indicator name, the indicator description and the indicator alias is calculated, and the similarity between the vector of the indicator information and the vectors corresponding to the indicator name, the indicator description and the indicator alias is calculated to obtain multiple indicator recall results; Sorting the indicator recall results based on the similarity, and determining the target retrieval indicator corresponding to each indicator information based on the sorting result; A search result is obtained from the indicator platform based on the target search indicator.

4. The method according to claim 3, characterized in that Acquiring search results from the indicator platform based on the target search indicator includes: Acquire production data associated with the target search indicator from the indicator platform as a search result of the indicator information; Alternatively, production data associated with the target retrieval indicator is obtained from the indicator platform, and calculation is performed based on the production data and the calculation logic to obtain a retrieval result of the indicator information.

5. The method according to claim 3, characterized in that: Also includes: Obtain historical conversation records of the first language model; The historical conversation records, the retrieval questions and the indicator recall results are input into the first large language model to sort the indicator recall results, and the first indicator recall result in the sorted results is determined as the target retrieval indicator.

6. The method according to claim 3, characterized in that: Based on the keyword and the keyword vector, a search and calculation is performed on a pre-built indicator platform to obtain a search result for each indicator information, including: For each indicator information, based on the search library, calculating the similarity between the indicator information and the indicator name, the indicator description and the indicator alias, and calculating the similarity between the vector of the indicator information and the vectors corresponding to the indicator name, the indicator description and the indicator alias; Determine an indicator recall result based on the similarity, a preset similarity threshold and a maximum recall quantity, input the indicator recall result and the search question into a predetermined rearrangement model, and return the first indicator recall result after rearrangement as a target search indicator; A search result is obtained from the indicator platform based on the target search indicator.

7. The method according to claim 3, characterized in that Based on the keyword and the keyword vector, a search and calculation is performed on a pre-built indicator platform to obtain a search result for each indicator information, including: For each indicator information, based on the search library, calculating the similarity between the indicator information and the indicator name, the indicator description and the indicator alias, and calculating the similarity between the vector of the indicator information and the vectors corresponding to the indicator name, the indicator description and the indicator alias; Determining an indicator recall result based on the similarity, a preset similarity threshold and a maximum recall quantity; Inputting the index recall result and the search question into the second largest language model to obtain a target search index; A search result is obtained from the indicator platform based on the target search indicator.

8. A data retrieval device based on an indicator platform, characterized in that: include: An indicator extraction module, used to obtain a search question input by a user, and extract keywords of the search question based on a pre-trained first language model; wherein the keywords include: time information and at least one indicator information; A vectorization module, used for performing vectorization processing on the keyword based on the fine-tuned vector model to obtain a vector of the keyword; A retrieval module is used to perform retrieval and calculation on a pre-built indicator platform based on the keyword and the vector of the keyword to obtain retrieval results of indicator information; wherein the retrieval results of the indicator information include: calculation results of a single production data or multiple production data; the indicator platform is connected to different data sources to obtain production data, and the indicator platform has multiple built-in calculation logics, and the calculation logic is used to calculate the production data.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

Images