Metering field information extraction method and system
By applying a large language model based on code style in the field of metrology, the problem of difficult to ensure the reliability of structured information in the field of metrology is solved, and effective classification of task text feature attributes and accurate output of structured information is achieved.
Patent Information
- Application Number
- CN202510197188.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-21
- Publication Date
- 2025-06-20
AI Technical Summary
The prior art is difficult to ensure the reliability of structured information during the information extraction process in the field of metrology, especially when there are fewer data samples.
Using a large language model based on code style, we use class definition generation modules, task text analysis modules, code logic conversion modules and structured information output modules to build a class definition set containing feature attributes, and convert them into task logic in the form of code, and finally output structured information.
Through this method, the feature attributes of the task text can be effectively classified, the accuracy and reliability of information extraction can be improved, and accurate structured information can be obtained even in the case of zero samples.
Smart Images
Figure CN120181076A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of metrology technology, and in particular, to a method and system for information extraction in the field of metrology. Background Art
[0002] Information Extraction (IE), as an important branch in the field of natural language processing, is dedicated to mining valuable structured information from unstructured text data. This information can be entities, relationships, events, etc., which form the basis for constructing knowledge graphs, supporting decision-making systems, and other intelligent applications. With the development of the Internet and the advent of the big data era, IE tasks have become increasingly crucial because more and more data exists in unstructured or semi-structured forms.
[0003] Currently, in the application of traditional IE methods to specific fields, such as metrology data, supervised learning dominates, which relies on a large amount of manually annotated data to train models. Such methods include, but are not limited to, Named Entity Recognition (NER), Relation Extraction (RE), and event extraction. In recent years, with the rise of deep learning, neural network-based language models, such as BERT (Bidirectional Encoder Representations from Transformers) and its variants, have significantly improved the performance of IE tasks. However, due to the small number of data samples in the metrology field, especially when it comes to specific structured information extraction tasks, it is difficult to ensure the accuracy of information extraction using the above methods.
[0004] Therefore, in the process of information extraction in the metrology field in the prior art, there is a problem that it is difficult to ensure the reliability of structured information. Summary of the Invention
[0005] In view of this, it is necessary to provide a method and system for information extraction in the metrology field to solve the problem that it is difficult to ensure the reliability of structured information in the process of information extraction in the prior art in the metrology field.
[0006] To solve the above problems, in a first aspect, the present invention provides a method for information extraction in the metrology field, including: Constructing a large language model based on code style, where the large language model based on code style includes a class definition generation module, a task text parsing module, a code logic transformation module, and a structured information output module; Input the task text in the metrology field into a large language model based on code style, and generate a set of class definitions containing the characteristic attributes of the task text according to the class definition generation module; Based on the task text parsing module, perform code transformation on the set of class definitions, and determine the code class definition related to the task text as the classification style descriptor; Based on the code logic transformation module, transform the classification style descriptor into the task logic in code form; Based on the structured information output module, output the structured information of the task text according to the task logic.
[0007] In a possible implementation, generating a set of class definitions containing the characteristic attributes of the task text according to the class definition generation module includes: Identify the characteristic attributes of the task text; Construct a set of class definitions containing the characteristic attributes through the chain of thought method.
[0008] In a possible implementation, identifying the characteristic attributes of the task text further includes: Obtain the data flow diagram of the task text; Obtain the semantic logic association relationship of the task text based on the data flow diagram; Extract the semantic logic features of the task text according to the semantic logic association relationship; Identify the characteristic attributes of the task text based on the semantic logic features.
[0009] In a possible implementation, after constructing a set of class definitions containing the characteristic attributes through the chain of thought method, it further includes: Obtain the metrology knowledge of the external knowledge base and determine the semantic features of the metrology knowledge; Iteratively update the class definitions based on the semantic features.
[0010] In a possible implementation, based on the task text parsing module, performing code transformation on the set of class definitions and determining the code class definition related to the task text as the classification style descriptor includes: Perform code transformation on each class definition in the set of class definitions one by one to obtain the corresponding descriptors; Annotate the descriptors with class comments, and generate multiple attributes and comments according to the class comment results; Regard the descriptors, class comment results, attributes, and comments as the element set of the class definition; Traverse and count the occurrence frequencies of the elements in the element set in the task text, and determine the descriptor corresponding to the highest occurrence frequency as the classification style descriptor of the task text.
[0011] In a possible implementation, the large language model based on code style further includes a retrieval enhancement module; The retrieval enhancement module is used to adjust the search strategy of the task text parsing module when traversing and counting the occurrence frequencies of elements in the statistical element set in the task text.
[0012] In a possible implementation manner, based on adjusting the search strategy of the task text parsing module when traversing and counting the occurrence frequencies of elements in the statistical element set in the task text, it includes: Obtain the key parameters of the task text according to the retrieval enhancement module; Determine the target search strategy according to the mapping relationship between the key parameters and the search strategy.
[0013] In a possible implementation manner, based on the code logic conversion module converting the classification style descriptor into the task logic in code form, it includes: Convert the task requirements of the task text into code logic to obtain the code framework of the task text; Based on the programming language to express the logical branches and iterative processes of the task text, determine the task logic of the task text.
[0014] In a possible implementation manner, after the structured information output module outputs the structured information of the task text based on the task logic, it further includes: Update the knowledge graph in the measurement field according to the structured information to obtain an updated knowledge graph including the task text.
[0015] In a second aspect, the present invention further provides an information extraction system in the measurement field, including a memory and a processor, wherein, The memory is used to store programs; The processor is coupled to the memory and is used to execute the programs stored in the memory to implement the steps in the above-mentioned information extraction method in the measurement field.
[0016] The beneficial effects of adopting the above embodiments are as follows: The present invention provides an information extraction method in the measurement field. By using the class definition generation module to perform class definition on the task text in the measurement field, the classification of the characteristic attributes of the task text is realized; by using the task text parsing module to perform correlation analysis on the task text and convert it into code form, a classification style descriptor is obtained; by using the code logic conversion module to summarize and sort out the logic of the classification style descriptor, in order to make full use of the reusability of the code, and then directly use the task logic to obtain the structured information of the task text based on zero samples, effectively ensuring the reliability of the structured information. Description of the Drawings
[0017] Figure 1 It is a schematic flowchart of an embodiment of the information extraction method in the measurement field provided by the present invention; Figure 2Schematic flowchart of an embodiment for obtaining a set of class definitions provided by the present invention; Figure 3 Schematic flowchart of an embodiment for identifying the characteristic attributes of a task text provided by the present invention; Figure 4 Schematic flowchart of an embodiment for determining a classification style descriptor provided by the present invention; Figure 5 Schematic flowchart of an embodiment for determining a target search strategy provided by the present invention; Figure 6 Block diagram of an embodiment of an information extraction system in the metrology field provided by the present invention. Detailed implementation manners
[0018] The preferred embodiments of the present invention will be specifically described below with reference to the accompanying drawings. The accompanying drawings form a part of the present invention and are used together with the embodiments of the present invention to explain the principles of the present invention, rather than to limit the scope of the present invention.
[0019] In order to solve the problem in the prior art that in the process of information extraction in the metrology field, it is difficult to ensure the reliability of the information extraction result due to the small number of metrology data samples, the present invention provides a metrology field information extraction method and system, which will be described in detail below.
[0020] As Figure 1 shown, Figure 1 Schematic flowchart of an embodiment of a metrology field information extraction method provided by the present invention, including: S101: Construct a large language model based on code style. The large language model based on code style includes a class definition generation module, a task text parsing module, a code logic transformation module, and a structured information output module; In some embodiments of the present invention, the large language model based on code style is essentially still a large language model. Specifically, the large language model based on code style can be any one of the GPT (Generative Pre-trained Transformer) series models, LLaMA (Large Language Model Meta AI) series models, BERT (Bidirectional Encoder Representations from Transformers) model, and T5 (Text-To-Text Transfer Transformer) model, and can be specifically selected and adjusted according to actual needs.
[0021] S102: Input the task text in the metrology field into a large language model based on code style, and generate a set of class definitions containing the characteristic attributes of the task text according to the class definition generation module; In some embodiments of the present invention, the task text in the metrology field is diverse and complex, specifically including unstructured text data, structured information, experimental reports or measurement records in metrology data, image and video data, sensor data, log data, simulation data, mixed data types, and so on. In particular, unstructured data refers to data that is not convenient to be represented by a two-dimensional logical table in a database, and usually includes various formats of office documents, text, pictures, XML, HTML, reports, image and audio / video information, etc. In the metrology field, unstructured text data may include: 1. Detailed descriptions and background information of experiments or measurements.
[0022] 2. Usage instructions, calibration records, and maintenance logs of equipment.
[0023] 3. Document materials of metrology standards and regulations.
[0024] These data usually need to be extracted and analyzed through technical means such as text mining and natural language processing.
[0025] In order to extract information in the metrology field in a diversified manner, this application mainly targets unstructured data, and through mining and analyzing irregular and inconveniently recognizable unstructured data, structured data that is convenient to sort out is obtained.
[0026] In some embodiments of the present invention, the task text in the metrology field refers to text content related to metrology activities, experiments, measurements, or data analysis, etc. This text content contains rich metrology data and descriptive information, which is the key to understanding and analyzing problems in the metrology field. Specifically, the task text corresponding to the metrology field includes but is not limited to experimental reports, measurement records, calibration certificates, data analysis reports, metrology standards, and technical specifications.
[0027] In a specific embodiment, a class definition is a comprehensive description and definition of a certain class of objects or entities. It covers the characteristic attributes, behaviors, and relationships with other classes that are common to the objects of this category.
[0028] Furthermore, the relationship between the characteristic attributes and the class definition includes: constitutive relationship, dependency relationship, dynamic adjustment, and application level.
[0029] Taking the "temperature" descriptor in measurement data as an example, its characteristic attributes may include units (such as degrees Celsius, Fahrenheit, etc.) and magnitude values (specific temperature values). These characteristic attributes together constitute the class definition of the "temperature" category. When we need to describe or process data related to temperature, these characteristic attributes will provide us with clear guidance and basis.
[0030] S103: Based on the task text parsing module, perform code transformation on the class definition set and determine the code class definition related to the task text as the classification style descriptor; The classification style descriptor refers to the descriptor code in code form obtained by transforming the task text into code form. It not only has the code style but also carries the data information of the task text, providing convenience for subsequent data processing, analysis, and storage.
[0031] S104: Based on the code logic transformation module, transform the classification style descriptor into the task logic in code form; S105: Based on the structured information output module, output the structured information of the task text according to the task logic.
[0032] In this embodiment, through the class definition generation module, the task text in the measurement field is class-defined, realizing the classification of the characteristic attributes of the task text; through the task text parsing module, the relevance analysis of the task text is carried out and transformed into code form to obtain the classification style descriptor; through the code logic transformation module, the logic of the classification style descriptor is summarized and sorted out to make full use of the reusability of the code, and then directly use the task logic to realize the structured information of the task text based on zero-sample acquisition, effectively ensuring the reliability of the structured information.
[0033] In some embodiments of the present invention, in S102, in order to generate a class definition set containing the characteristic attributes of the task text according to the class definition generation module, such as Figure 2 shown Figure 2 is a schematic flowchart of an embodiment of obtaining a class definition set provided by the present invention, including: S201: Identify the characteristic attributes of the task text; In some embodiments of the present invention, the characteristic attribute refers to the inherent or external characteristics possessed by the instances or categories in the data set, which are the inherent attributes of the task text. These characteristics can be numerical, such as height and weight, or categorical, such as gender and species.
[0034] In some embodiments of the present invention, in order to identify the characteristic attributes of the task text, it is necessary to clarify the goal and background of the task text, preview the data set, analyze the data attributes, determine the characteristic attributes, verify the characteristic attributes, and document the characteristic attributes, etc., to ensure that the characteristic attributes can accurately describe the data.
[0035] In some embodiments of the present invention, in order to specifically illustrate how to identify the characteristic attributes of task texts, such as Figure 3 as shown Figure 3 is a schematic flowchart of an embodiment for identifying the characteristic attributes of task texts provided by the present invention, including: S301: Obtain the data flow diagram of the task text; S302: Obtain the semantic logical association relationship of the task text based on the data flow diagram; In some embodiments of the present invention, the data flow diagram (abbreviated as DFD) is the main expression tool of the structured system analysis method. From the perspective of data transfer and processing, it graphically expresses the logical functions of the system, the logical flow of data within the system, and the logical transformation process. It marks the logical input and logical output of a system, as well as the processing required to convert the logical input into the logical output. Since it only reflects the logical functions that the system must complete, the data flow diagram is a functional model.
[0036] In this embodiment, by directly capturing the semantic logical association relationship of the task text through the data flow diagram, the internal parameter relationship of the task text is sorted out from the logical level, which is beneficial to capturing the correlation between task texts, so as to improve the reliability of the model architecture during zero-shot learning.
[0037] S303: Extract the semantic logical features of the task text according to the semantic logical association relationship; In this embodiment, by capturing the key features in the semantic logical association relationship, the semantic logical features that can represent the internal logic of the task text are extracted, realizing the feature extraction of the task text from the logical level, and can assist the large language model based on code style to more accurately capture the semantic features of the task text.
[0038] S304: Identify the characteristic attributes of the task text based on the semantic logical features.
[0039] It should be noted that semantic logical features can capture the deep meaning and context relationship in the data, so as to more accurately identify the characteristic attributes of the data.
[0040] In this embodiment, by using semantic logical features to assist in identifying the characteristic attributes of task texts, it can reduce data processing errors caused by misunderstanding or misjudgment, improve the accuracy of data recognition, enhance the consistency and comprehensibility of data, and facilitate subsequent zero-shot learning assisted by large language models based on code style.
[0041] S202: Construct a set of class definitions containing characteristic attributes through the chain of thought method.
[0042] In a specific embodiment, to ensure the integrity of the class definition, after constructing a class definition including characteristic attributes based on the task text corresponding to the metrology field, it is also necessary to obtain the metrology knowledge of the external knowledge base and determine the semantic features of the metrology knowledge; then, the class definition is iteratively updated based on the semantic features.
[0043] In this embodiment, by using the semantic features of the metrology knowledge in the external knowledge base as the basis for determining the class definition, the scope of the class definition is effectively expanded, avoiding the problem that the requirements of the task text cannot be met subsequently.
[0044] In some embodiments of the present invention, in S103, in order to perform code transformation on the class definition set based on the task text parsing module and determine the code class definition related to the task text as the classification style descriptor, such as Figure 4 shown, Figure 4 is a schematic flowchart of an embodiment for determining the classification style descriptor provided by the present invention, including: S401: Perform code transformation on each class definition in the class definition set one by one to obtain the corresponding descriptor; S402: Perform class annotation on the descriptor and generate multiple attributes and annotations according to the class annotation results; S403: Define a set of elements including the descriptor, class annotation results, attributes, and annotations as the elements of the class definition; S404: Traverse and count the occurrence frequencies of the elements in the element set in the task text, and determine the descriptor corresponding to the highest occurrence frequency as the classification style descriptor of the task text.
[0045] In some embodiments of the present invention, the generated code class definition adapts to the specific task by analyzing the objects in the task text, ensuring that the meaning of the descriptor is closely aligned with the task requirements. For each descriptor, multiple class annotations are first generated, then multiple attributes and annotations are generated for each class annotation, and a class including the original descriptor, class annotation, and these attributes and annotations is defined. Then, a counter is initialized to record the number of object occurrences, and a set is formed including the descriptor, class annotation, and attributes and annotations. Subsequently, all task texts are traversed, the occurrence times of the elements in this set in each task text are counted, and the counter is updated. The fitness calculation is based on the total sum of the counter, which reflects the relevance or effectiveness of the class definition in describing the task text. Finally, the class definition with the highest fitness is selected and added to the classification style descriptor set. Repeat this process until all descriptors are processed, and finally return the classification style descriptor set. In this way, the descriptor that can best represent the task characteristics can be identified, ensuring that its meaning matches the task requirements. The specific process is as follows: (1) Feature attribute frequency calculation: Evaluate the relevance of each class definition to the task text, and perform preliminary screening by calculating the occurrence frequency of feature attributes in the task text.
[0046] (2) Fitness evaluation: Conduct a comprehensive evaluation of candidate class definitions, and select the one with the highest fitness as the final classification style descriptor.
[0047] (3) Task alignment: Ensure that the finally selected class definition can best reflect the task requirements and improve the accuracy of information extraction.
[0048] In this embodiment, by performing multi-element annotation on each descriptor, the class annotation results, attributes, and annotations of the descriptor are obtained, and by defining an element set that combines the descriptor, class annotation results, attributes, and annotations as the elements of the class definition, the association between the class definition and its elements is realized; then, taking the task text in the metrology field as the search library, the descriptor corresponding to the highest occurrence frequency value is determined as the classification style descriptor of the class definition, realizing the relationship mapping between the class definition and the descriptor, thereby obtaining a complete and reliable set of classification style descriptors.
[0049] In some embodiments of the present invention, in order to improve the efficiency when traversing and counting the occurrence frequency of elements in the element set in the task text, the large language model based on code style further includes a retrieval enhancement module; The retrieval enhancement module is used to adjust the search strategy of the task text parsing module when traversing and counting the occurrence frequency of elements in the element set in the task text.
[0050] Specifically, as Figure 5 shown, Figure 5 is a schematic flowchart of an embodiment of the target search strategy provided by the present invention, including: S501: Obtain the key parameters of the task text according to the retrieval enhancement module; It should be noted that the key parameters can not only guide the system to select the most appropriate indexing method and search algorithm, but also ensure that the retrieval strategy can closely fit the actual needs of the task. In the metrology field, the key parameters mainly include the following aspects: 1. Term type and domain specificity Term type: This involves whether the terms used in the query are general terms or domain-specific professional terms. The system needs to identify these terms and adjust the search strategy according to their types to more accurately match relevant information.
[0051] Domain specificity: For metrological data processing, the system needs to identify whether the query is targeted at a specific domain (such as physics, chemistry, engineering, etc.). Different domains may have different metrological standards and term systems, so the system needs to adjust the search strategy according to the domain characteristics.
[0052] 2. Hierarchical Relationship of Structured Information Information Hierarchy: Measurement data often contains complex structured information, such as experimental design, measurement steps, data results, etc. The system needs to identify the hierarchical relationship of this information and adjust the search strategy according to the hierarchical structure to understand and retrieve relevant data more deeply.
[0053] Association Rules: In the processing of measurement data, there may be specific association rules or dependencies between data. The system needs to identify these rules and consider these relationships during the search process to ensure the accuracy and integrity of the retrieval results.
[0054] 3. Measurement Standards and Experimental Conditions Measurement Standards: Different measurement tasks may follow different measurement standards or specifications. The system needs to identify these standards and adjust the search strategy according to the standards to ensure that the retrieved information meets specific measurement requirements.
[0055] Experimental Conditions: Experimental conditions (such as temperature, pressure, humidity, etc.) have an important impact on measurement results. The system needs to identify the experimental conditions mentioned in the query and adjust the search strategy according to these conditions to find data and information related to specific experimental conditions.
[0056] 4. Data Formats and Sources Data Formats: Measurement data may be stored in various formats, such as text, tables, images, etc. The system needs to identify the data format and select appropriate indexing methods and search algorithms according to the format.
[0057] Data Sources: Different data sources may provide information of different qualities and types. The system needs to identify the data sources mentioned in the query and adjust the search strategy according to the reliability, authority, and relevance of the sources.
[0058] S502: Determine the target search strategy according to the mapping relationship between key parameters and search strategies.
[0059] It should be noted that the search strategy includes indexing methods and search algorithms.
[0060] Among them, the indexing method is a technology used to improve the efficiency of database queries. By creating an index structure of data, the retrieval speed of data can be accelerated. Common indexing methods include Primary Key Index, Unique Index, Index / Non-unique Index, Full-Text Index, and Composite Index / Combined Index / Multi-column Index.
[0061] A search algorithm is a method that utilizes the high performance of a computer to purposefully exhaust some or all possible cases in the solution space of a problem, thereby finding the solution to the problem. Common search algorithms include linear search (Linear Search), binary search (Binary Search), depth-first search (Depth-First Search, DFS), breadth-first search (Breadth-First Search, BFS), A* search algorithm, hash function, and index search algorithm.
[0062] Key parameters play a crucial role in dynamically adjusting the search strategy to optimize the information retrieval process. The system needs to comprehensively consider these parameters and select the most suitable indexing method and search algorithm according to the actual situation to ensure the accuracy and efficiency of the retrieval results.
[0063] In a specific embodiment, according to the requirements of a specific task, the system can dynamically adjust the search strategy and optimize the information retrieval process. For example, when the task requirements include terms in a specific field or complex structured information, the system will establish a mapping relationship between the parameters and the search strategy to achieve dynamic adjustment of the search strategy. The specific process is as follows: First, the system analyzes the key parameters in the task requirements, such as the type of terms, the hierarchical relationship of structured information, etc.; then, it selects the most suitable indexing method and search algorithm according to these parameters. For example, when facing queries in different fields or types, the module will automatically select the most suitable indexing method and search algorithm. This not only improves the retrieval efficiency but also enhances the system's ability to handle diverse tasks. For metrological data processing, this flexibility means that the system can flexibly adjust the retrieval strategy according to specific measurement standards or experimental conditions to ensure the acquisition of the most relevant information.
[0064] In summary, in this embodiment, with reference to the key parameters of the task text, the search strategy of the task text parsing module is adjusted from the perspective of adjusting the search strategy to improve the reliability of the task requirements.
[0065] In some embodiments of the present invention, after determining the classification style descriptor of the task text, in order to achieve data anchoring in the metrological field, that is, to limit the classification style descriptor in the metrological field to avoid data disorder problems, it is also necessary to anchor the classification style descriptor in a specific task text; Among them, the specific task text refers to the text related to the metrological field.
[0066] In this embodiment, by delimiting the classification style descriptor with the metrological field as the boundary, it is possible to reduce the influence of data from other similar fields on the data in this field and improve the pertinence of the classification style descriptor.
[0067] It should be noted that the two principles of MSTM (Master of Science Technology Management for Experienced Professionals) are as follows: Principle 1: If the type of an object is unknown, it can be identified by its characteristic attributes.
[0068] Principle 2: If the meaning of a descriptor is unclear, it can be inferred from the text related to its task.
[0069] Based on the two principles of MSTM, by making full use of the characteristic attributes of descriptors to enhance the recognition of substantive text objects, and anchoring the meaning of descriptors in specific task texts to achieve a zero-shot learning solution, it helps to better identify objects and their characteristic attributes, as well as infer the meaning of descriptors, and improve the recognition ability of substantive text objects in zero-shot learning scenarios.
[0070] In some embodiments of the present invention, in S104, in order to convert the classification style descriptor into a task logic in code form based on the code logic conversion module, first, the task requirements of the task text are converted into code logic to obtain the code framework of the task text; then, based on the programming language, the logical branches and iterative processes of the task text are expressed to determine the task logic of the task text.
[0071] In some embodiments of the present invention, specifically, convert the descriptor class definition into a task logic in code form, use conditional statements, loops, and function calls to express complex task logic, and implement automatic entity or concept recognition and classification, automatically execute tasks and generate accurate results. The specific operations include: (1) Task logic conversion: Precisely convert the task requirements described in natural language into code logic to ensure that the goals, inputs, and expected outputs of the task are clearly defined. By analyzing the background information and specific requirements of the task, the system generates the corresponding code framework, which covers all necessary variable declarations, function definitions, and control structures.
[0072] (2) Complex logic expression: After clarifying the basic framework of the task, use the syntax structure of the programming language (such as if-else statements, for / while loops) to express the logical branches and iterative processes in the task. It not only involves simple conditional judgments and loop executions, but also includes the ability to handle multi-layer nested logics, exception situations, and concurrent operations. The system can accurately simulate the dynamic behavior of the task to ensure that each logical branch can be correctly processed, thereby improving the reliability and accuracy of task execution.
[0073] (3) Automated execution: By instantiating and reasoning, entities or concepts in the text are identified and classified, and results that meet the task requirements are automatically generated.
[0074] In some embodiments of the present invention, after obtaining the structured information, in order to better guide the metrology work, it is also necessary to update the knowledge graph in the metrology field according to the structured information to obtain an updated knowledge graph containing the task text.
[0075] It should be noted that a knowledge graph is a series of various different graphs showing the development process and structural relationships of knowledge, using visualization technology to describe knowledge resources and their carriers, and mining, analyzing, constructing, drawing, and displaying knowledge and their interconnections.
[0076] To better implement the information extraction method in the metrology field in the embodiments of the present invention, correspondingly, the embodiments of the present invention also provide a metrology field information extraction system, as Figure 6 shown Figure 6 is a structural block diagram of an embodiment of the metrology field information extraction system provided by the present invention. The metrology field information extraction system 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only some components of the metrology field information extraction system 600 are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.
[0077] In some embodiments, the processor 601 may be a central processing unit (CPU), a microprocessor, or other data processing chips, and is used to run the program code stored in the memory 602 or process data, such as the metrology field information extraction method in the present invention.
[0078] In some embodiments, the processor 601 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 601 may be local or remote. In some embodiments, the processor 601 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination of the above.
[0079] The memory 602 can be an internal storage unit of the metrology field information extraction system 600 in some embodiments, such as the hard disk or memory of the metrology field information extraction system 600. The memory 602 can also be an external storage device of the metrology field information extraction system 600 in other embodiments, such as a plug-in hard disk equipped on the metrology field information extraction system 600, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.
[0080] Furthermore, the memory 602 can also include both the internal storage unit of the metrology field information extraction system 600 and the external storage device. The memory 602 is used to store the application software and various types of data for installing the metrology field information extraction system 600.
[0081] The display 603 can be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. in some embodiments. The display 603 is used to display the information in the metrology field information extraction system 600 and to display a visual user interface. The components 601-603 of the metrology field information extraction system 600 communicate with each other through the system bus.
[0082] In one embodiment, when the processor 601 executes the metrology field information extraction program in the memory 602, the following steps can be implemented: Construct a large language model based on code style, and the large language model based on code style includes a class definition generation module, a task text parsing module, a code logic transformation module, and a structured information output module; Input the task text in the metrology field into the large language model based on code style, and generate a set of class definitions containing the characteristic attributes of the task text according to the class definition generation module; Based on the task text parsing module, perform code transformation on the set of class definitions, and determine the code class definition related to the task text as the classification style descriptor; Based on the code logic transformation module, transform the classification style descriptor into the task logic in code form; Based on the structured information output module, output the structured information of the task text according to the task logic.
[0083] It should be understood that when the processor 601 executes the metrology field information extraction program in the memory 602, in addition to the above functions, other functions can also be implemented. For specific details, please refer to the description of the corresponding method embodiments above.
[0084] Further, the embodiments of the present invention do not specifically limit the type of the information extraction system 600 in the measurement field. The information extraction system 600 in the measurement field may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop, etc. Exemplary embodiments of the portable electronic device include, but are not limited to, portable electronic devices equipped with IOS, android, microsoft or other operating systems. The above portable electronic device may also be other portable electronic devices, such as a laptop with a touch-sensitive surface (such as a touch panel). It should also be understood that in some other embodiments of the present invention, the information extraction system 600 in the measurement field may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (such as a touch panel).
[0085] Correspondingly, an embodiment of the present application also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the programs or instructions are executed by a processor, the steps or functions in the measurement field information extraction method provided by the above method embodiments can be implemented.
[0086] Those skilled in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory, etc.
[0087] The above has introduced in detail the measurement field information extraction method and system provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.
Claims
1. A method for extracting information in the field of metrology, characterized in that: include: Constructing a large language model based on code style, wherein the large language model based on code style includes a class definition generation module, a task text parsing module, a code logic conversion module and a structured information output module; Inputting the task text in the metrology field into the large language model based on the code style, and generating a class definition set including the characteristic attributes of the task text according to the class definition generation module; Performing code conversion on the class definition set based on the task text parsing module, and determining the code class definition related to the task text as a classification style descriptor; The classification style descriptor is converted into a task logic in code form based on the code logic conversion module; The structured information of the task text is output based on the structured information output module according to the task logic.
2. The method for extracting measurement domain information according to claim 1, characterized in that: The step of generating a class definition set including characteristic attributes of the task text according to the class definition generation module includes: Identifying characteristic attributes of the task text; A class definition set containing the characteristic attributes is constructed through chain thinking.
3. The method for extracting measurement domain information according to claim 2, characterized in that: The identifying characteristic attributes of the task text also includes: Obtain a data flow diagram of the task text; Acquire the semantic logical association relationship of the task text based on the data flow graph; Extracting semantic logical features of the task text according to the semantic logical association relationship; The feature attributes of the task text are identified based on the semantic logical features.
4. The method for extracting measurement domain information according to claim 2, characterized in that: After constructing a class definition set containing the characteristic attributes through chain thinking, it also includes: Acquire metrological knowledge from an external knowledge base and determine semantic features of the metrological knowledge; The class definition is iteratively updated based on the semantic features.
5. The method for extracting measurement domain information according to claim 1, characterized in that: The code conversion of the class definition set based on the task text parsing module and determining the code class definition related to the task text as a classification style descriptor includes: Converting multiple class definitions of the class definition set into codes one by one to obtain corresponding descriptors; Performing class annotation on the descriptor, and generating a plurality of attributes and annotations according to the class annotation result; Treat the descriptor, the class annotation result, the attributes and the annotation as a set of elements of the class definition; The occurrence frequencies of the elements in the element set in the task text are traversed and counted, and the descriptor corresponding to the highest occurrence frequency value is determined as the classification style descriptor of the task text.
6. The method for extracting measurement domain information according to claim 5, characterized in that: The large language model based on code style also includes a retrieval enhancement module; The retrieval enhancement module is used to adjust the search strategy of the task text parsing module when traversing and counting the occurrence frequency of elements in the element set in the task text.
7. The method for extracting measurement domain information according to claim 6, characterized in that: The step of adjusting the search strategy of the task text parsing module when traversing and counting the occurrence frequency of elements in the element set in the task text includes: Acquire key parameters of the task text according to the retrieval enhancement module; The target search strategy is determined according to the mapping relationship between the key parameters and the search strategy.
8. The method for extracting measurement field information according to claim 1, characterized in that: The task logic of converting the classification style descriptor into code form based on the code logic conversion module includes: Convert the task requirements of the task text into code logic to obtain a code framework of the task text; The task logic of the task text is determined based on the logical branches and iterative process of expressing the task text in a programming language.
9. The method for extracting measurement domain information according to claim 1, characterized in that: After the structured information output module outputs the structured information of the task text according to the task logic, the method further includes: The knowledge graph of the metrology field is updated according to the structured information to obtain an updated knowledge graph including the task text.
10. A metrology field information extraction system, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for extracting metrology field information as described in any one of claims 1 to 9 above.