Natural resource-based spatio-temporal data element value mining method and device

Abstract

The embodiment of the invention provides a natural resource-based spatio-temporal data element value mining method and device, and solves the problem of insufficient spatio-temporal data mining. The method comprises the following steps: performing arrangement and paragraph segmentation on data elements to form a vector database; training and describing the vector database to form a data set knowledge base; receiving user input data, and analyzing a scene required by a user based on the user input data; querying the data set knowledge base based on the user input data to determine data and a calculation method corresponding to a scene required by the user; and calling a workflow interface, obtaining the corresponding data and calculation method, and recommending a processing result to a user.

Description

Technical Field

[0001] The present invention relates to the technical field of data element value mining, and in particular to a method and device for mining the value of natural resource spatiotemporal data elements. Background Art

[0002] Spatiotemporal data has the characteristics of wide sources and diverse types. In the field of natural resources, spatiotemporal data has the typical characteristics of huge data volume, numerous nodes, and complex business relationships. According to the business field, it can be divided into natural resource survey and monitoring data, national land space planning business data, use control and analysis and evaluation business fields; according to the data source, it can be divided into: aerospace / aeronautical collection data, annual change survey data, planning data generated by different business fields; according to the data space form, it can be divided into: point, line, surface, complex surface and other types of data; according to the data storage method, it can be divided into: relational database storage, file geographic database storage, file storage and other types. Although with the improvement of digital governance capabilities, the organization and management of natural resource data are becoming more and more standardized, there is still a problem of insufficient mining of data information value. The main problems can be summarized as follows: how to mine explicit and implicit information and value from spatiotemporal data; how to perform fusion analysis and correlation analysis on multiple data; the national / provincial volume data is huge and cannot be calculated efficiently; when a user gives a piece of data, how to quickly give the applicable scenario of this data and realize automatic recommendation from data to scenario. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a method and device for mining the value of spatiotemporal data elements of natural resources, which solves the problem of insufficient spatiotemporal data mining.

[0004] In a first aspect, an embodiment of the present invention provides a method for mining the value of natural resource spatiotemporal data elements, comprising: Arrange and segment data elements to form a vector database; Training and describing the vector database to form a data set knowledge base; Receiving user input data, and analyzing the user's required scenario based on the user input data; Querying the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user; The workflow interface is called to obtain the corresponding data and calculation method, and the processing results are recommended to the user.

[0005] In one embodiment, the data elements are sorted and segmented into paragraphs to form a vector database, including: Classifying the data elements according to different business scenarios; The classified data elements are subjected to named entity recognition and text similarity calculation. Through contextual semantic relationship understanding and text differential weight learning, the data elements are segmented from articles to paragraphs and from paragraphs to sentences to form a vector database.

[0006] In one embodiment, after the data elements are sorted and segmented into paragraphs to form a vector database, the method further includes: Expand the number of data elements in the vector database; Set model parameters, train the model for multiple rounds based on the self-supervised training method and the data in the vector database, and fine-tune local parameters during the training process; After each training session, the model effect is evaluated until the accuracy is within the expected range, thus completing the preparation of the training data set.

[0007] In one embodiment, the number of data elements in the expanded vector database includes: Expand data elements based on the business scenarios manually sorted out in the data elements; Calculate the business scenario similarity of the expanded data elements, and screen the data elements based on the obtained similarity to generate representative business scenarios in other fields; Generate step-by-step business scenario processing flow based on platform tools; Convert the step-by-step business scenario processing flow and overall business description into structured text.

[0008] In one embodiment, the training and describing of the vector database to form a data set knowledge base includes: based on retrieval enhancement and generation technology, training and describing the vector database, matching the data set according to text content and user intent, retrieving matching text fragments from the knowledge base, and parsing the data set fragments and retrieval results through multi-way recall mode configuration and performing semantic reordering to form a data set knowledge base.

[0009] In one embodiment, the querying of the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user includes: matching the user input data with the data in the data set knowledge base through semantic similarity, and obtaining the data and calculation method corresponding to the scenario required by the user based on the model logical reasoning capability.

[0010] In one embodiment, before calling the workflow interface, it includes: reading various types of natural resource data and parsing the data structure, defining nodes and branches for the parsed data to construct a workflow, and forming a workflow interface.

[0011] In a second aspect, an embodiment of the present invention provides a device for mining the value of natural resource spatiotemporal data elements, comprising: A sorting and segmentation unit, used for sorting data elements and segmenting them into paragraphs to form a vector database; A training and description unit, used for training and describing the vector database to form a data set knowledge base; A receiving and analyzing unit, configured to receive user input data and analyze the user's desired scenario based on the user input data; A query and comparison unit, used to query the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user; The calling and outputting unit is used to call the workflow interface, obtain the corresponding data and calculation method, and recommend the processing result to the user.

[0012] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the steps of the above-mentioned method when executed by a processor.

[0013] In a fourth aspect, an embodiment of the present invention provides a computer program product, comprising a computer program, which implements the steps of the method described in the claim above when executed by a processor.

[0014] The embodiment of the present invention provides a method and device for mining the value of spatiotemporal data elements of natural resources, which organizes and segments data elements to form a vector database; trains and describes the vector database to form a data set knowledge base; receives user input data, analyzes the user's required scenario based on the user input data; queries the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the user's required scenario; calls the workflow interface to obtain the corresponding data and calculation method, and recommends the processing results to the user. This method has the following advantages over the prior art: (1) It has the ability to self-perceive and self-learn data. Through the continuous accumulation of knowledge base, fine-tuning training and parameter adjustment of large models, the system will continuously optimize its understanding of each piece of data; (2) When a new piece of data is input, the big model can quickly provide information and scenario recommendations on multiple dimensions of the data; (3) For natural resource spatiotemporal data, a knowledge material library that can be accumulated is formed, including business data related to natural resource survey and monitoring and national land space planning, forming a complete, reusable, and self-growing element data set. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1Shown is a flow chart of a method for value mining based on natural resource spatiotemporal data elements provided by an embodiment of the present invention.

[0016] Figure 2 Shown is a flow chart of natural language processing and text segmentation provided by an embodiment of the present invention.

[0017] Figure 3 Shown is a flow chart of forming an access interface provided by an embodiment of the present invention.

[0018] Figure 4 The figure shows a flow chart of organizing data sets and knowledge bases based on Dify provided by an embodiment of the present invention.

[0019] Figure 5 Shown is a flowchart of preparing a training data set provided by an embodiment of the present invention.

[0020] Figure 6 The figure shows a flow chart of expanding the number of data elements in a vector database provided by one embodiment of the present invention.

[0021] Figure 7 The flowchart of the process of combining natural resource spatiotemporal data elements with a large model and the process of building a workflow is shown in an embodiment of the present invention. Figure 8 The figure shows a flow chart of a device for mining the value of natural resource spatiotemporal data elements provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, and to fully understand and implement how the present disclosure applies technical means to solve technical problems and achieve the corresponding technical effects, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. The embodiments of the present disclosure and the various features in the embodiments can be combined with each other without conflict, and the technical solutions formed are all within the scope of protection of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0024] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] In response to the above problems, users are increasingly concerned about the value mining content of data elements. How to quickly realize the explicit and implicit information expression in a piece of data is the core issue that needs to be considered in this invention. In response to this problem, at the technical level, it is considered to use the knowledge base and fine-tuning capabilities of the large model to parse and train the time information, spatial information, and field structure of the data, organize and fine-tune the data application scenarios, application models, and analysis processes, and finally realize interactive analysis and value mining. When the user gives a piece of data on the front-end interface, the system will intelligently give the value of the data in which dimensions, which scenarios it can be used for, which business problems it can solve, and which business system construction it supports.

[0026] The explicit business model is a deterministic business scenario with clear input and output, calculation process and calculation indicators, such as common natural resource business: survey and monitoring business analysis, urban development boundary potential assessment, cultivated land inflow and outflow statistical analysis, high-standard farmland and stable cultivated land analysis and evaluation, land acquisition data and the third national land survey data overlay analysis, the province's basic farmland data and slope data overlay assignment calculation; the second category is implicit value mining needs, including the relationship analysis between population activities and urban land, comprehensive analysis and evaluation of urban land and traffic conditions, mobile phone signaling data and urban planning business combined analysis, land use data and agricultural meteorological data combined analysis, mobile phone signaling data and urban planning business combined analysis, industrial land type and environmental governance relationship analysis, this type of demand belongs to data-driven scenario construction.

[0027] In the traditional application model, business needs drive scenario construction, which includes the data and analysis tools used to build a business model. In other words, platform construction provides a kind of analysis and calculation tools and capabilities, and system construction provides a kind of expression and solution to business problems, but both lack value discovery from the perspective of data. When different categories of data are cross-displayed and organized, how to mine value from the data dimension and realize intelligent scenario recommendation is a core topic that needs to be considered.

[0028] Therefore, the ultimate goal of this invention is to realize the intelligent linkage of data, scenarios, tools and models, so that users can quickly mine the value of elements from both the data dimension and the business dimension.

[0029] In one embodiment of the present invention, a method for mining the value of natural resource spatiotemporal data elements is provided. Figure 1 As shown, the method includes: Step 01: Arrange and segment data elements into paragraphs to form a vector database. Classify the data elements according to different business scenarios; perform named entity recognition and text similarity calculation on the classified data elements, and segment the data elements from articles to paragraphs and from paragraphs to sentences through contextual semantic relationship understanding and text differential weight learning to form a vector database.

[0030] refer to Figure 2 As shown, the description, purpose, and common business scenarios of the data elements themselves are sorted into text materials, and then the sorted data elements are input into the Bert model. Through the natural language recognition and text segmentation capabilities of the Bert model, the incoming business scenario knowledge base materials, text materials, etc. are sorted and segmented into paragraphs to form a vector database. Among them, BERT (Bidirectional Encoder Representations from Transformers) is the full name of "Bidirectional Encoder Representations" or simply "Bidirectional Transformer Model", which is a pre-trained language model based on the Transformer architecture.

[0031] Classify according to different business scenarios, such as natural resource survey and monitoring scenarios, national land space planning business scenarios, and implementation supervision analysis and evaluation business scenarios. Each scenario covers five parts: business demand description, input natural resource data, calculation process and analysis tool processing relationship, analysis parameters and spatial reference, and analysis result data output. Through the MLM (Masked Language Model) learning method of the Bert model, named entity recognition and text similarity calculation are realized. Through contextual semantic relationship understanding and text differential weight learning, the splitting from articles to paragraphs and from paragraphs to sentences is realized to form a vector database.

[0032] Step 02: Train and describe the vector database to form a dataset knowledge base. Based on retrieval enhancement and generation technology, train and describe the vector database, match the dataset according to the text content and user intent, retrieve matching text fragments from the knowledge base, parse the dataset fragments and retrieval results through multi-way recall mode configuration, and perform semantic reordering to form a dataset knowledge base.

[0033] Through the Dify large model application development platform, three capabilities are realized, including workflow services, persistent dialogue capabilities, vector retrieval capabilities, and intelligent question-answering services. Among them, Dify is an open source large language model (LLM) application development platform. Through Dify's built-in RAG (Retrieval Augmented Generation, a technical method that combines retrieval and generation) retrieval and generation technology, the vector database generated in the previous step is trained and described, the data set is matched according to the text content and user intent, and the matching text fragments are retrieved from the knowledge base. Through the re-ranking multi-way recall mode configuration, the data set fragments and retrieval results are semantically re-ordered, thereby realizing the data set knowledge base generation capability and setting a high-quality retrieval priority mode in the knowledge base.

[0034] Step 03: Receive user input data, and analyze the user's required scenario based on the user input data.

[0035] Specifically, scenarios are divided according to the daily work text materials and national standard materials of enterprises and units in the input data, and each scenario includes information such as which data and which analysis tools are needed. Among them, the business scenario is known.

[0036] Step 04: Based on the user input data, query the data set knowledge base to determine the data and calculation method corresponding to the scenario required by the user. The user input data and the data in the data set knowledge base are matched by semantic similarity, and the data and calculation method corresponding to the scenario required by the user are obtained based on the model logical reasoning ability.

[0037] When a user enters a piece of data for description and scenario recommendation, the system will first compare and search the vectors based on the knowledge base organized by Dify, match them by semantic similarity, and call on the logical reasoning capabilities of the large model to provide a sufficiently complete answer.

[0038] Before calling the workflow interface, the method further includes: reading various types of natural resource data and parsing the data structure, defining nodes and branches for the parsed data to construct a workflow and form a workflow interface.

[0039] as follows Figure 3 As shown in the figure, for various types of natural resource data (such as file geographic databases, relational databases, online services, distributed file systems, etc.), the data structure is read and parsed through the program, and the workflow is built through Dify to achieve multi-dimensional metadata information parsing of the data, covering data name, data description, data volume and scale, spatial projection, geometric type, time information and other dimensional content. By defining nodes and branches, it is combined into a workflow, and finally forms an access interface for system integration development. Reference diagram of data set and knowledge base organization based on Dify Figure 4 shown.

[0040] Step 05: Call the workflow interface to obtain the corresponding data and calculation method, and recommend the processing results to the user.

[0041] Call the workflow interface to generate intelligent scenario recommendations. Through interactive question-and-answer format, a profound data description is given for the input natural resource spatiotemporal data, including the data's spatiotemporal characteristics, geometric characteristics, structural characteristics, etc. It can also be reorganized and processed according to the data's field conditions and fragment descriptions in the knowledge base to form scenario-based recommendations for data applications, such as which scenarios the data can be applied to, which businesses it can support, and which problems it can solve, forming a closed loop at the data application level.

[0042] When the user inputs two or more sets of data, based on the semantic understanding and self-learning ability of the large model, it can automatically recommend correlation analysis, such as analysis of the management relationship between population activities and urban land, analysis of the balance between urban planning and personnel's work-residence relationship, comprehensive analysis of land use planning and agricultural meteorological data, analysis of the correlation relationship between industrial land types and environmental governance, etc. Each scenario is formed through comprehensive analysis of multiple data elements.

[0043] Users of this system organize a text description of their current work or project and input it into the model. The model analyzes the scenario required by the user according to the natural language analysis in step 01, queries the vector database, and determines the data and calculation method corresponding to the scenario. Then, the workflow interface is called to obtain data and calculation methods, and the processing results are recommended to the user.

[0044] In one embodiment of the present invention, the QWEN large model API is called to generate a fine-tuning data set, wherein the API model is a framework for building and designing an application programming interface (API). It is used through parameter configuration and fine-tuning. The fine-tuning method uses lora (Low-Rank Adaptation, a technology for optimizing deep learning models). Through self-supervised training, the number of training rounds, maximum gradient norm, maximum number of samples and other parameters of the model are set, multiple rounds of training are performed, and local parameters are fine-tuned. Finally, the model effect is evaluated until the accuracy is within the expected range, and the training data set preparation is completed. After the data elements are sorted and segmented to form a vector database, as described above, Figure 5 As shown, it also includes: Step 011: Expand the number of data elements in the vector database. Figure 6 As shown, this step includes: Step 0111: Expand the data elements based on the business scenarios manually sorted in the data elements; Step 0112: Calculate the business scenario similarity of the expanded data elements, and screen the data elements based on the obtained similarity to generate representative business scenarios in other fields; Step 0113: Generate a step-by-step business scenario processing flow based on the platform tool; Step 0114: Convert the step-by-step business scenario processing flow and the overall business description into structured text.

[0045] Step 012: setting model parameters, performing multiple rounds of model training based on the self-supervised training method and the data in the vector database, and fine-tuning local parameters during the training process; Step 013: After each training, the model effect is evaluated until the accuracy is within the expected range, and the training is completed to complete the preparation of the training data set.

[0046] In one embodiment of the present invention, the following steps are required to convert a small amount of manually sorted business scenario data into a large amount of data required for fine-tuning a large model: generalizing the business scenario, generating specific processing steps, and formatting processing steps.

[0047] Generalized business scenarios refer to: by letting the large model refer to a small number of manually sorted business scenarios to draw inferences from one example, and by using the BERT model to calculate and screen the similarity of business scenarios, representative business scenarios in other fields are automatically generated. Through this step, we have obtained a large number of business scenarios in other fields described in popular natural language.

[0048] Generating specific processing steps means: generating a step-by-step business scenario processing flow based on platform tools by allowing the big model to refer to the manually organized business scenario processing steps and the data analysis and processing tools available on the platform.

[0049] The formatting processing step refers to: by letting the large model refer to the manually compiled JSON format reference text, the step-by-step processing flow and the overall business description are converted into a structured JSON text, which includes five contents: business requirement description, input natural resource data, calculation process and analysis tool processing relationship, analysis parameters and spatial reference, and analysis result data output.

[0050] In one embodiment of the present invention, the generalized business scenario is divided into the following steps: 1. Generate program initialization: Read configuration files, seed data, technical documents and prompt word templates, and initialize running parameters.

[0051] 2. Text Generation: The generate function is called in multiple threads to concatenate reference text and prompt words, and generate instruction text through the OpenAI API.

[0052] Gather the generated instruction texts into a general library.

[0053] 3. Similarity calculation: Multi-threaded calculation of the similarity between the newly generated instruction text and the retained text to avoid repeated calculation.

[0054] Use the BERT model to calculate semantic similarity and return the result to the main thread.

[0055] 4. Text filtering: Use sparse matrices to store similarity results, filter out text sequences with similarity below a threshold, and retain high-quality instructions.

[0056] 5. Save the results: Clean the stop words in the text and save the generated instruction text and similarity matrix.

[0057] The prompt word structure of the generalized business scenario is as follows: 1. Reference documents: Reference information sources for generating business scenarios 2. Task Description: Clear task objectives: Generate a specified number of instruction texts around technical texts.

[0058] Emphasize output format: output line by line, and strictly prohibit output of explanatory text.

[0059] 3. Specific requirements: Try not to repeat verbs.

[0060] Diversify tone and approach (e.g., question, imperative).

[0061] Diversified task types (such as open generation, classification, editing, etc.).

[0062] Output strictly in the example format.

[0063] No text or explanatory information irrelevant to the business scenario is output.

[0064] 4. Example reference: Several sample instructions are provided as a reference for the format of generating new instructions.

[0065] Sample instructions should cover a variety of tones, approaches, and task types.

[0066] In one embodiment of the present invention, the specific processing steps are divided into the following steps: 1. Program initialization: Read the configuration file, initialize parameters, and load breakpoint data (if any).

[0067] 2. Text Generation: The generate function is called in multiple threads to concatenate the reference text and the instruction text, and the reply text is generated through the OpenAI API.

[0068] 3. Task Management: Track the completion of the generated tasks and determine whether the generated results meet the format requirements.

[0069] 4. Save and breakpoints: Save the reply text that meets the requirements into the dictionary, save the generation progress regularly, and support breakpoint resumption.

[0070] The structure of the prompt words for generating specific processing steps is as follows: 1. Reference Documents: Reference information sources for task generation 2. Task Description: Clarify task objectives: analyze business scenarios and model them.

[0071] Emphasize the output format: strictly follow the {"instruction":"","input":"","output":""} structure, and each field cannot be empty.

[0072] 3. Specific requirements: Only operators defined in the operator list can be used; functions cannot be made up.

[0073] The operator name must contain both Chinese and English names in the format of [name (alias)].

[0074] Each step can only contain one function and multiple functions cannot be combined.

[0075] The steps need to be reasonable and logical.

[0076] No explanatory information is output.

[0077] If no explicit instructions are given, readShapefile is used by default.

[0078] If no explicit instructions are given, the default value is [RDSWriterToolFunction (write to resource pool)].

[0079] Reasonably select operators based on business scenarios, such as spatial overlay, field filtering, area calculation, etc.

[0080] 4. Output example: The example should demonstrate the complete analysis process with clear steps and reasonable logic.

[0081] In one embodiment of the present invention, the formatting process is divided into the following steps: 1. Program initialization: Read the configuration file, initialize parameters, and load breakpoint data (if any).

[0082] 2. Text Generation: The generate function is called in multiple threads to concatenate the reference text and the instruction text, and the reply text is generated through the OpenAI API.

[0083] 3. Task Management: Track the completion of the generated tasks and determine whether the generated results meet the format requirements.

[0084] 4. Save and breakpoints: Save the reply text that meets the requirements into the dictionary, save the generation progress regularly, and support breakpoint resumption.

[0085] The prompt word structure of the formatting processing step is as follows: 1. Mission Description: Provide a clear task description, that is, you need to convert the provided original format text into JSON structured data in the target format.

[0086] 2. Step by step guide: Analysis Task: Instruct how to identify the operation name and description of each step from the original text.

[0087] Information filling: Accurately fill in the corresponding fields of the target JSON format (such as target, inputdata, description, handlesteps, arguments, and outputdata) based on the information in the original text.

[0088] Parameter extraction: Ensure that the corresponding description is accurately extracted or constructed according to the step number in the arguments section, and include path information (if applicable).

[0089] Output definition: In the outputdata section, clearly indicate the data type of the final output and its save path.

[0090] 3. Example provided: A specific original format text example (such as "Land classification statistics for law enforcement in satellite images") and an example of the expected target JSON format (such as "Potential space for urban development boundaries") are given for reference.

[0091] The specific contents of the prompt words for the formatting processing step are as follows: Please convert the original format text provided below into the target format JSON structured data. During the conversion process, make sure to follow the following steps: 1. Analyze the tasks in the original text and identify the operation name and description of each step.

[0092] 2. Fill in the fields corresponding to the target format based on the information provided in the original text, including but not limited to `target`, `inputdata`, `description`, `handlesteps`, `arguments`, and `outputdata`.

[0093] 3. For the `arguments` section, make sure to accurately extract or construct the corresponding description based on the step number and include path information if provided.

[0094] 4. In the `outputdata` section, clearly indicate the data type of the final output (such as shp, Excel) and its save path.

[0095] In one embodiment of the present invention, reference Figure 7 As shown, this invention revolves around natural resource data elements, starting from the data, to mine and express the implicit value information of the data itself. Relying on the generalization and self-learning capabilities of the large language model (LLM), the knowledge base organization and large model training of multi-source data are realized, and the ability of data resource scenario design, scenario prediction, and scenario recommendation is realized through the LLM interface. The key point of this invention is to start from the data elements, and to start and train from the data structure, type, and tense, breaking the previous one-size-fits-all data source access and management methods, but to perceive and trace the data elements, and through the combination with the large model, to realize the data set knowledge base organization and fine-tuning, split the data description, form contextual semantic relationships, and display the front end through the workflow mode, and realize the value mining of data elements in an interactive way. The summary is as follows: 1) Starting from the data elements, sort out the geometric information, spatial form, field structure, application scenarios, and scenarios that can be integrated; 2) Knowledge base and fine-tuning processing: separate text fields through multiple rounds of training and fine-tuning, using the generalization learning ability of the large model to improve the quality of the data set; 3) Interactive intelligent question and answer recommendation: Through the interactive web page, each piece of input data is described and explained, outlining the data structure, volume scale, application scenarios, and business applications of multiple data fusion analysis.

[0096] This embodiment provides a device 100 based on the value mining of natural resource spatiotemporal data elements, such as Figure 8 As shown, the device includes: an arrangement and segmentation unit 10, a training and description unit 20, a receiving and analysis unit 30, a query and comparison unit 40 and a call and output unit 50. The sorting and segmenting unit 10 is used to sort the data elements and segment them into paragraphs to form a vector database; The training and description unit 20 is used to train and describe the vector database to form a data set knowledge base; The receiving and analyzing unit 30 is used to receive user input data and analyze the user's required scenario based on the user input data; The query and comparison unit 40 is used to query the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user; The calling and outputting unit 50 is used to call the workflow interface, obtain the corresponding data and calculation method, and recommend the processing result to the user.

[0097] Furthermore, the sorting and segmentation unit 10 is used to classify the data elements according to different business scenarios; perform named entity recognition and text similarity calculation on the classified data elements, and segment the data elements from articles to paragraphs and from paragraphs to sentences through contextual semantic relationship understanding and text differential weight learning to form a vector database.

[0098] Furthermore, the sorting and segmentation unit 10 is also used to expand the number of data elements in the vector database; set model parameters, perform multiple rounds of model training based on the self-supervised training method and the data in the vector database, and fine-tune local parameters during the training process; evaluate the model effect after each training until the accuracy is within the expected range, thereby completing the preparation of the training data set.

[0099] Furthermore, the sorting and segmentation unit 10 is also used to expand the data elements based on the business scenarios manually sorted in the data elements; calculate the business scenario similarity of the expanded data elements, and screen the data elements based on the obtained similarity to generate representative business scenarios in other fields; generate a step-by-step business scenario processing flow based on platform tools; and convert the step-by-step business scenario processing flow and the overall business description into structured text.

[0100] Furthermore, the training and description unit 20 is used to train and describe the vector database based on retrieval enhancement and generation technology, match the data set according to the text content and user intention, retrieve matching text fragments from the knowledge base, and parse the data set fragments and retrieval results through a multi-way recall mode configuration and perform semantic reordering to form a data set knowledge base.

[0101] Furthermore, the receiving and analyzing unit 30 is used to match the user input data with the data in the data set knowledge base through semantic similarity, and at the same time obtain the data and calculation method corresponding to the scenario required by the user based on the model logical reasoning ability.

[0102] Furthermore, the query and comparison unit 40 is used to read various types of natural resource data and parse the data structure, define nodes and branches for the parsed data to construct a workflow, and form a workflow interface.

[0103] On the basis of the above embodiments, this embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the above embodiments.

[0104] In some implementations of this embodiment, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the method described in the above embodiment are implemented.

[0105] In some implementations of this embodiment, a computer program product is provided, including a computer program / instructions, characterized in that when the computer program is executed by a processor, the steps of the method described in the above embodiment are implemented.

[0106] The processor may include, but is not limited to, one or more processors or microprocessors. Each processor may be an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor or other electronic components to execute the method in the above embodiments. The computer-readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, and the computer-readable storage medium may include but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (e.g., hard disk, floppy disk, solid-state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0107] The computer-readable storage medium may also store at least one computer executable program / instruction, which may be, for example, a computer-readable instruction. The computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The computer-readable storage medium may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions stored on the computer-readable storage medium, the various methods described above may be performed.

[0108] In addition, the computer device may also include (but not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (eg, keyboard, mouse, speaker, etc.), etc.

[0109] The processor may communicate with external devices via an I / O bus via a wired or wireless network.

[0110] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product / computer program product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0111] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0112] It should be noted that in the present disclosure, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element limited by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0113] Although the embodiments disclosed in the present disclosure are as above, the above contents are only embodiments adopted for facilitating the understanding of the present disclosure and are not intended to limit the present disclosure. Any technician in the technical field to which the present disclosure belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present disclosure, but the scope of patent protection of the present disclosure shall still be subject to the scope defined in the attached claims.

Claims

1. A method for mining the value of natural resource spatiotemporal data elements, characterized in that: include: Arrange and segment data elements to form a vector database; Training and describing the vector database to form a data set knowledge base; Receiving user input data, and analyzing the user's required scenario based on the user input data; Querying the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user; The workflow interface is called to obtain the corresponding data and calculation method, and the processing results are recommended to the user.

2. The method for mining the value of natural resource spatiotemporal data elements according to claim 1 is characterized in that: The data elements are sorted and segmented into paragraphs to form a vector database, including: Classifying the data elements according to different business scenarios; The classified data elements are subjected to named entity recognition and text similarity calculation. Through contextual semantic relationship understanding and text differential weight learning, the data elements are segmented from articles to paragraphs and from paragraphs to sentences to form a vector database.

3. The method for mining the value of natural resource spatiotemporal data elements according to claim 1 is characterized in that: After the data elements are sorted and segmented into paragraphs to form a vector database, the method further includes: Expand the number of data elements in the vector database; Set model parameters, train the model for multiple rounds based on the self-supervised training method and the data in the vector database, and fine-tune local parameters during the training process; After each training session, the model effect is evaluated until the accuracy is within the expected range, thus completing the preparation of the training data set.

4. The method for mining the value of natural resource spatiotemporal data elements according to claim 3 is characterized in that: The number of data elements in the extended vector database includes: Expand data elements based on the business scenarios manually sorted out in the data elements; Calculate the business scenario similarity of the expanded data elements, and screen the data elements based on the obtained similarity to generate representative business scenarios in other fields; Generate step-by-step business scenario processing flow based on platform tools; Convert the step-by-step business scenario processing flow and overall business description into structured text.

5. The method for mining the value of natural resource spatiotemporal data elements according to claim 1 is characterized in that: The training and describing of the vector database to form a data set knowledge base includes: based on retrieval enhancement and generation technology, training and describing the vector database, matching the data set according to text content and user intention, retrieving matching text fragments from the knowledge base, and parsing the data set fragments and retrieval results through multi-way recall mode configuration and performing semantic reordering to form a data set knowledge base.

6. The method for mining the value of natural resource spatiotemporal data elements according to claim 1 is characterized in that: The querying of the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user includes: matching the user input data with the data in the data set knowledge base through semantic similarity, and obtaining the data and calculation method corresponding to the scenario required by the user based on the model logical reasoning ability.

7. The method for mining the value of natural resource spatiotemporal data elements according to claim 1 is characterized in that: Before calling the workflow interface, it includes: reading various types of natural resource data and parsing the data structure, defining nodes and branches for the parsed data to construct a workflow, and forming a workflow interface.

8. A device for mining the value of natural resource spatiotemporal data elements, characterized in that: include: A sorting and segmentation unit, used for sorting data elements and segmenting them into paragraphs to form a vector database; A training and description unit, used for training and describing the vector database to form a data set knowledge base; A receiving and analyzing unit, configured to receive user input data and analyze the user's desired scenario based on the user input data; A query and comparison unit, used to query the data set knowledge base based on the user input data to determine the data and calculation method corresponding to the scenario required by the user; The calling and outputting unit is used to call the workflow interface, obtain the corresponding data and calculation method, and recommend the processing result to the user.

9. 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 implemented.

10. A computer program product, comprising a computer program, 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 implemented.

Images