Data semantic perception and adaptive parsing method and system for non-networking devices

Abstract

The application provides a data semantic perception and adaptive analysis method and system for non-networking equipment, and relates to the technical field of intelligent data processing of experimental equipment. The existing technology is difficult to realize automatic collection of multi-source data for non-networking heterogeneous experimental equipment with various interface types, great communication protocol differences and no network communication capability, lacks semantic understanding and unified expression of heterogeneous data content, and cannot automatically generate and continuously evolve the analysis rules when new equipment is connected. The application simultaneously realizes automatic collection of non-networking heterogeneous experimental equipment data, semantic-level unified expression of heterogeneous parameters, and automatic generation and continuous evolution of analysis rules without networking modification by constructing an integrated intelligent analysis framework combining device semantic knowledge base and data semantic perception, and introduces a knowledge-driven verification mechanism to ensure the reliability of the analysis results.

Description

Technical Field

[0001] This invention relates to the field of intelligent data processing technology for experimental equipment, and in particular to a method and system for data semantic perception and adaptive parsing for non-networked devices. Background Technology

[0002] Laboratory testing and inspection institutions are equipped with a large number of non-networked experimental devices, such as pH meters, digital vernier calipers, electronic balances, and chromatographs. These devices generate a large amount of experimental data during daily testing and inspection. However, these devices generally suffer from problems such as diverse interface types, large differences in communication protocols, heterogeneous data output formats, and the lack of network communication capabilities. This results in experimental data relying on manual transcription, sorting, and entry into electronic lab notebooks by testing personnel for a long time. This not only leads to low data collection efficiency but also easily introduces transcription errors and human bias, severely restricting the utilization efficiency and reliability of data traceability. Although some new devices have begun to support standard communication protocols or have been integrated into electronic lab notebook systems, many older models of devices—such as pH meters and electronic balances that only have serial port or USB output functions, or even those that only display measurement results on a screen—still cannot directly access data via network protocols, becoming a prominent bottleneck in laboratory information technology development. Therefore, achieving automatic data collection from non-networked devices without requiring network upgrades to existing equipment is an urgent problem to be solved in this field.

[0003] To address the aforementioned issues, existing technologies include methods based on general data acquisition systems. These systems connect multiple devices via serial communication, USB communication, or network communication, with a central server collecting data from each device. Related technical specifications have been widely applied in industrial and some laboratory equipment interoperability practices, leading to the exploration of information models and modeling standards around analytical and testing equipment. Furthermore, research has proposed standardized data processing workflows to automatically download, merge, process, and transmit data from multiple instruments, verifying the feasibility of automatically acquiring and uniformly processing data from multiple heterogeneous instruments. However, these general data acquisition solutions primarily target devices that support standard communication protocols or are integrated into electronic lab notebooks. For a large number of older, non-networked devices with only serial or USB output capabilities, or even those that only display data on a screen, automatic data acquisition remains difficult. Simultaneously, these solutions mainly handle data aggregation, transmission, and format conversion, lacking semantic understanding of the data content. They cannot perform unified parameter identification and semantic alignment for heterogeneous data output from different devices. Different devices may use different parameter names and unit systems for the same physical quantity, making it difficult to directly integrate and utilize the parsed data. Therefore, how to further endow the system with the ability to understand the semantic content of data on the basis of realizing automatic data collection from non-networked devices, and realize the unified semantic expression of heterogeneous data, is another technical problem that urgently needs to be solved in this field.

[0004] Furthermore, rule-based parsing methods rely on manual configuration of regular expressions and field separators for data parsing. However, rules must be rewritten when new devices are connected, leading to cumbersome configuration and high maintenance costs, hindering the automatic generation and continuous evolution of parsing rules. Semantic integration-based methods utilize ontology models and semantic annotation techniques to uniformly represent device data, but they primarily target networked or specific single devices, exhibiting limited adaptability to multi-source data from non-networked heterogeneous devices and lacking dynamic adaptive parsing capabilities. Some studies have proposed methods for extracting instrument screen measurement values ​​based on image processing techniques, but these only address single devices and cannot constitute a complete multi-source heterogeneous data parsing solution. Existing methods have significant shortcomings in three aspects: automatic data acquisition from non-networked devices, semantic understanding of heterogeneous data, and automatic generation of parsing rules. Currently, a complete technical solution capable of simultaneously addressing these three issues is lacking. This invention constructs an integrated intelligent parsing framework that combines device semantic modeling and data semantic perception. This framework enables automatic acquisition, semantic-level understanding, and adaptive parsing of data from non-networked heterogeneous experimental devices without requiring network modifications, effectively solving many of the problems existing in the aforementioned technologies. Summary of the Invention

[0005] Existing technologies have significant shortcomings in automatic data acquisition from offline devices, semantic understanding of heterogeneous data, and automatic generation of parsing rules, making it difficult to simultaneously achieve automatic acquisition, unified semantic expression, and adaptive parsing of data from offline heterogeneous experimental devices. Therefore, this invention first proposes a data semantic perception and adaptive parsing method for offline devices, specifically including the following steps: Step S1: Construct a semantic knowledge base for non-networked devices, collect basic information about experimental devices, and establish a semantic knowledge base for device category tags, experimental parameters, and device parsing rules. Step S2: Based on the device category label and device resolution rule base, automatically identify the interface type of the experimental device, and obtain multi-source experimental data from non-networked devices and cache them uniformly based on the identification results; Step S3: Based on the experimental parameter semantic knowledge base, perform semantic annotation on the cached experimental data, identify experimental parameters, units and dimensions, and map the identified parameters to the standard semantic space to obtain the parsed data with semantic tags. Step S4: Based on the semantic tags and device features in the data before parsing, call the corresponding parsing rules in the device parsing rule base to perform structured parsing on the experimental data to obtain standardized data; for unknown devices not recorded in the device parsing rule base, transfer the parsing rules based on the feature similarity between the unknown device and the known device and adaptively adjust them to generate parsing rules adapted to the unknown device. Step S5: Based on the parameter association constraint relationship in the experimental parameter semantic knowledge base, perform consistency verification on the standardized data, evaluate the credibility of the parsing results based on the verification results, and mark the parsing results with credibility lower than the preset threshold as pending verification. Step S6: Output standardized experimental data that has been verified and evaluated for reliability.

[0006] Furthermore, in step S1, the experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information, and the correlation constraints between parameters; the equipment parsing rule base includes data format templates, field separators, and numerical extraction rules for various types of equipment; and the equipment category labels include at least pH meters, digital vernier calipers, electronic balances, and chromatographs.

[0007] Furthermore, in step S2, the interface types include serial port, USB, Ethernet, and image acquisition interface; for devices that support serial port or USB communication, identification and connection are achieved by reading the device descriptor and communication parameters; for devices that only display data on the screen, screen images are captured by a camera and data is extracted using OCR technology; the unified cache includes device number, acquisition time, original data fields, and data length information.

[0008] Furthermore, in step S3, semantic annotation is implemented based on a neural network model, and the annotation results include parameter category labels, unit labels, and experimental condition labels; mapping to the standard semantic space is achieved by calculating the semantic similarity between the target device parameters and the standard parameters in the experimental parameter semantic knowledge base, and parameters with similarity below a preset threshold are marked as requiring manual confirmation.

[0009] Furthermore, in step S4, the device features are constructed based on the semantic tags and device communication features in the data before parsing; the parsing rules include at least a data format template, field separators, numerical extraction positions, and unit conversion coefficients; the migration parsing rules are as follows: calculate the feature similarity between the unknown device and each known device in the device parsing rule library, select the parsing rule of the most similar known device as the initial template, and fine-tune the field separators and numerical extraction positions in the initial template according to the actual output format of the unknown device.

[0010] Furthermore, in step S5, the consistency verification includes unit consistency verification and parameter logical relationship verification; unit consistency verification is to uniformly convert the same physical quantity in different unit systems to the standard unit system; parameter logical relationship verification is to verify whether the parameters in the analysis result meet the preset physical constraint relationship in the experimental parameter semantic knowledge base; credibility is a quantitative score obtained by weighted comprehensive evaluation of unit consistency, parameter logical relationship and numerical range rationality.

[0011] Furthermore, in step S6, the standardized experimental data includes at least the unique device identifier, experimental timestamp, parameter name, standard value, standard unit, and confidence score, and is output in JSON or CSV format and stored in the experimental database.

[0012] According to another aspect of the present invention, a data semantic awareness and adaptive parsing system for non-networked devices is provided, applied to the above-described method, comprising: The equipment semantic knowledge base construction module is used to collect basic information about experimental equipment and establish equipment category tags, experimental parameter semantic knowledge bases, and equipment parsing rule bases. The experimental data acquisition module is used to automatically identify the interface type of experimental equipment based on equipment category tags and equipment parsing rule base, and to obtain multi-source experimental data from non-networked experimental equipment and cache it uniformly based on the identification results; The data semantic awareness module is used to semantically annotate the cached experimental data based on the experimental parameter semantic knowledge base, identify experimental parameters, units and dimensions, and map the identified parameters to the standard semantic space to obtain the parsed data with semantic tags. The adaptive parsing module is used to perform structured parsing of experimental data by calling the corresponding parsing rules in the device parsing rule library based on the semantic tags and device features in the data before parsing, and obtain standardized data; for unknown devices not recorded in the device parsing rule library, the parsing rules are transferred based on the feature similarity between the unknown device and the known device and adaptively adjusted to generate parsing rules adapted to the unknown device. The knowledge-driven verification module is used to perform consistency verification on standardized data based on the parameter association constraints in the experimental parameter semantic knowledge base, evaluate the credibility of the parsing results based on the verification results, and mark the parsing results with credibility below a preset threshold as pending verification. The standardized data output module is used to output standardized experimental data that has been verified and evaluated for reliability.

[0013] Furthermore, in the device semantic knowledge base construction module, the experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information, and the correlation constraints between parameters; the device parsing rule base includes data format templates, field separators, and numerical extraction rules for various types of devices; in the experimental data acquisition module, the interface types include serial port, USB, Ethernet, and image acquisition interface. For devices that support serial port or USB communication, identification and connection are achieved by reading device descriptors and communication parameters; for devices that only display data on the screen, screen images are acquired through a camera and data is extracted using OCR technology.

[0014] Furthermore, the data semantic perception module performs semantic annotation on experimental data based on neural networks, and realizes the mapping of parameters to the standard semantic space by calculating the semantic similarity between parameters and standard parameters; in the adaptive parsing module, the transfer parsing rule selects the parsing rule of the most similar known device as the initial template and fine-tunes it according to the actual output format of the unknown device; in the knowledge-driven verification module, consistency verification includes unit consistency verification and parameter logical relationship verification, and the credibility is a weighted comprehensive evaluation of unit consistency, parameter logical relationship and the rationality of numerical range; the data output by the standardized data output module includes at least device identifier, timestamp, parameter name, standard value, standard unit and credibility score.

[0015] Furthermore, the dynamic Bayesian network module includes a structure construction submodule, a conditional probability table initialization submodule, a forward inference submodule, a backward inference submodule, and a cross-time-slice inference submodule. The structure construction submodule is used to map causal paths in the knowledge graph into directed acyclic graphs. The conditional probability table initialization submodule is used to initialize conditional probability tables based on historical case statistics and expert fuzzy evaluation. The forward inference submodule is used to calculate system-level risk probabilities using the belief propagation algorithm. The backward inference submodule is used to calculate the posterior probability contribution of each root node to the top node, identify key root causes, and verify the consistency of causal paths based on the knowledge graph. The cross-time-slice inference submodule is used to continuously update the risk probability time series using a sliding window approach.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: Firstly, this invention proposes a device knowledge-driven adaptive parsing mechanism. By establishing a device semantic knowledge base and a parsing rule base, it achieves automatic device type identification. When encountering an unknown device, it can match the closest device model based on device characteristics and automatically adjust the parsing rules, enabling continuous learning and autonomous evolution of the parsing algorithm. This solves the technical problems of existing technologies where new device access requires manual writing of parsing rules, cumbersome configuration, and high maintenance costs. This significantly improves parsing accuracy and greatly reduces cumulative configuration time, effectively lowering the manual and time costs of new device access. Secondly, this paper also provides a data semantic perception method for non-networked heterogeneous devices. Through semantic annotation, it maps heterogeneous parameters output by different devices to a unified standard semantic space, realizing the transformation of data from field-level parsing to semantic-level understanding. This solves the technical problems of existing technologies lacking semantic understanding capabilities of data content and being unable to perform unified parameter recognition and semantic alignment of heterogeneous data output by different devices. It can maintain high semantic recognition accuracy under different data acquisition conditions, effectively handle parameter semantic recognition tasks under diverse data formats, and achieve automatic semantic-level understanding of data output from heterogeneous devices. Third, a data credibility verification method constrained by semantic knowledge is constructed. This method incorporates experimental domain knowledge, physical constraints of parameters, and experimental process logic into the data parsing process. It performs multi-level consistency verification on the parsing results and establishes a credibility assessment model, enabling credibility assessment and automatic error correction of experimental data. This solves the technical problems of insufficient credibility guarantee in parsing results and difficulty in automatically identifying and correcting abnormal data in existing technologies. It can also automatically identify and correct various data anomalies such as inconsistent units and OCR recognition errors. While completing data parsing, it ensures the semantic consistency, unit uniformity, and credibility of the data, providing a reliable data foundation for subsequent intelligent analysis of experimental data. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 Flowchart of a data semantic awareness and adaptive parsing method for non-networked devices; Figure 2 A graph comparing the parsing accuracy performance of other parsing methods on different device types; Figure 3 A graph comparing the accuracy of parametric semantic recognition with other semantic recognition methods under different data acquisition conditions; Figure 4 A comparison chart showing the time required to resolve rule configurations with other configuration methods under varying numbers of new devices connected; Figure 5 A diagram illustrating standardized data output and reliability assessment on non-networked heterogeneous experimental equipment; Figure 6 This is a schematic diagram of the data semantic perception and adaptive parsing system structure for non-networked devices in the embodiment. Figure 7 This is a schematic diagram of a data semantic perception and adaptive parsing system for non-networked devices. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0020] The specific embodiments of the present invention will be described below.

[0021] Existing technologies struggle to automatically collect multi-source data from heterogeneous, non-networked experimental devices with diverse interface types, differing communication protocols, and lacking network communication capabilities. They also lack semantic understanding and unified representation of heterogeneous data content, and require manual configuration of parsing rules when new devices are connected, failing to automatically generate and continuously evolve rules. This invention addresses this by constructing an integrated intelligent parsing framework that combines a device semantic knowledge base with data semantic perception. Without requiring network modifications, it simultaneously achieves automatic data collection from non-networked heterogeneous devices, unified semantic representation of heterogeneous parameters, and automatic generation and continuous evolution of parsing rules. Furthermore, a knowledge-driven verification mechanism ensures the reliability of the parsing results.

[0022] Example 1 like Figure 1 The flowchart illustrates a data semantic perception and adaptive parsing method for non-networked devices. This invention proposes a data semantic perception and adaptive parsing method for non-networked devices, which specifically includes the following steps: Step S1: Construct a semantic knowledge base for non-networked devices, collect basic information on experimental equipment, and establish equipment category tags, experimental parameter semantic knowledge bases, and equipment parsing rule bases. The experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information, and the relationships and constraints between parameters; the equipment parsing rule base includes data format templates, field separators, and numerical extraction rules for each type of equipment; and the equipment category tags include at least pH meters, digital vernier calipers, electronic balances, and chromatographs.

[0023] Specifically, the basic information of each experimental device was first collected, including device model, manufacturer, communication interface type, data output format, and historical analysis experience. Analysis rule templates for 10 common types of experimental equipment were collected, covering pH meters, digital vernier calipers, electronic balances, universal testing machines, high-performance liquid chromatographs, gas chromatographs-mass spectrometers, UV-Vis spectrophotometers, rapid solvent extractors, water bath shakers, and eight-basket ovens.

[0024] Based on this, equipment category tags are established to identify different models and types of experimental equipment. A semantic knowledge base for experimental parameters is built, including parameter names, physical units, dimensional information, and the relationships and constraints between parameters. For example, pH meter output parameters include pH value, temperature, and electrode slope; electronic balance output parameters include mass value, calibration status, and units; and high-performance liquid chromatograph output parameters include concentration, retention time, and units. The semantic knowledge base also stores the relationships and constraints between parameters, such as the compensation relationship between pH value and temperature, and the relationship between mass value and calibration status. A device parsing rule base is established, including data format templates, field separators, and numerical extraction rules for various types of equipment. This provides a unified knowledge foundation for the entire parsing framework, providing knowledge support for subsequent data acquisition, semantic awareness, adaptive parsing, and knowledge-driven verification.

[0025] Step S2: Based on the device category label and device resolution rule base, the interface type of the experimental equipment is automatically identified. Multi-source experimental data is then obtained from non-networked experimental devices and cached uniformly based on the identification results. Interface types include serial port, USB, Ethernet, and image acquisition interfaces. For devices supporting serial port or USB communication, identification and connection are achieved by reading the device descriptor and communication parameters. For devices that only display data on the screen, screen images are captured via a camera, and data is extracted using OCR technology. The uniform cache includes device number, acquisition time, original data fields, and data length information.

[0026] Specifically, the interface type of the experimental equipment is first automatically identified based on device category tags and a parsing rule base. For devices supporting serial or USB communication, device identification and connection are achieved by reading the device descriptor and communication parameters, including baud rate, data bits, stop bits, and parity bits. For older devices that only display data on the screen, a high-definition camera is used to periodically capture screen images, and the experimental data from the screen is extracted using Optical Character Recognition (OCR) technology.

[0027] Then, based on device communication characteristics, output format, and historical parsing experience, the system automatically selects the appropriate data acquisition method. The collected multi-source experimental data is cached in a unified format, including device number, acquisition time, original data fields, and data length information. This enables automatic acquisition of data from non-networked devices, providing a complete data foundation for subsequent semantic perception and parsing.

[0028] Step S3: Based on the experimental parameter semantic knowledge base, semantic annotation is performed on the cached experimental data to identify experimental parameters, units, and dimensions. The identified parameters are then mapped to a standard semantic space to obtain pre-parsing data with semantic labels. Semantic annotation is implemented using a neural network model, and the annotation results include parameter category labels, unit labels, and experimental condition labels. Mapping to the standard semantic space is achieved by calculating the semantic similarity between the target device parameters and the standard parameters in the experimental parameter semantic knowledge base. Parameters with similarity below a preset threshold are marked as requiring manual confirmation.

[0029] Specifically, semantic processing is performed on the cached data based on the experimental parameter semantic knowledge base. First, the parameter names, units, timestamps, experimental conditions, and measurement results in the experimental data are semantically labeled. The semantic labeling model adopts a lightweight neural network architecture based on Transformer, containing a 3-layer encoder and a 2-layer fully connected classification head. The output dimension corresponds to the number of standard parameter categories in the semantic knowledge base. The semantic labeling process is represented as follows: ; in, S is the semantic annotation model determined by the model parameters θ, x is the original experimental data field, θ is the semantic annotation model parameter, and S is the semantic annotation result vector, which includes parameter category labels, unit labels, and experimental condition labels.

[0030] The annotation results are then mapped to a standard semantic space. By calculating the semantic similarity between the target device parameters and the standard parameters in the knowledge base, a unified expression of parameters across different devices is achieved. The semantic similarity calculation is represented as follows: ; Among them, P i Let P be the i-th experimental parameter to be semantically mapped in the target device. j Let v be the j-th standard parameter in the semantic knowledge base. i and v j P respectively i and P j The corresponding semantic embedding vector, This indicates the calculation of cosine similarity. Indicates parameter P i With parameter P j The semantic similarity between them. Parameters with similarity below a preset threshold are marked as requiring manual confirmation.

[0031] For example, a pH meter outputs raw data as pH=7.02, T=25.0°C. First, pH is identified as the parameter name label, 7.02 as the pH count value, T as the temperature parameter label, 25.0 as the temperature value, and °C as the unit label. Then, these identification results are mapped to a standard semantic space to obtain a unified semantic expression. By mapping heterogeneous data output from different devices to a unified semantic space, the transformation from field-level parsing to semantic-level understanding of data is achieved, eliminating semantic heterogeneity between different devices.

[0032] Step S4: Based on the semantic tags and device features in the pre-parsing data, the corresponding parsing rules in the device parsing rule base are called to perform structured parsing of the experimental data to obtain standardized data. For unknown devices not recorded in the device parsing rule base, the parsing rules are transferred and adaptively adjusted based on the feature similarity between the unknown device and known devices to generate parsing rules adapted to the unknown devices. The device features are based on the semantic tags and device communication features in the pre-parsing data. The parsing rules include at least a data format template, field separators, numerical extraction positions, and unit conversion coefficients. The transferred parsing rules are as follows: calculate the feature similarity between the unknown device and each known device in the device parsing rule base, select the parsing rule of the most similar known device as the initial template, and fine-tune the field separators and numerical extraction positions in the initial template according to the actual output format of the unknown device.

[0033] Specifically, adaptive parsing is performed based on semantic labels and device features in the pre-parsing data. Device type determination is based on similarity matching between the device feature vector and known device feature vectors in the knowledge base. The determination process is expressed as follows: ; in, The device type determination result for unknown devices. For the feature vector of the unknown device, Let k be the feature vector of the known device of the kth class in the knowledge base. This is a function for calculating the similarity of device feature vectors.

[0034] For known devices, corresponding parsing rules are automatically generated based on device type and historical parsing experience, including data format templates, field separators, numerical extraction positions, and unit conversion coefficients. For unknown devices not recorded in the knowledge base, parsing is completed through transfer learning: the feature similarity between the unknown device and known devices is calculated, the parsing rule of the most similar known device is selected as the initial template, and then the field separators and numerical extraction positions are fine-tuned according to the actual output format of the unknown device, thereby automatically generating parsing rules adapted to the device.

[0035] For example, when the system encounters a new model of pH meter for the first time, it finds through device feature matching that it has the highest similarity to the existing pH-001 model. Then, the parsing rules of pH-001 are used as the initial template, and the field separators and value extraction positions are fine-tuned according to the output format of the new model of pH meter, so that parsing rules adapted to the device can be automatically generated.

[0036] The final output is a unified standardized representation of heterogeneous experimental data. The standardized data format includes device number, experiment time, parameter name, standard value, standard unit, and confidence level. Automatic generation and continuous evolution of analytical rules are implemented.

[0037] Step S5: Based on the parameter association constraints in the experimental parameter semantic knowledge base, perform consistency verification on the standardized data. Evaluate the credibility of the parsing results based on the verification results, and mark parsing results with credibility below a preset threshold as requiring further review. Consistency verification includes unit consistency verification and parameter logical relationship verification. Unit consistency verification involves converting the same physical quantity in different unit systems to the standard unit system. Parameter logical relationship verification verifies whether the parameters in the parsing results satisfy the preset physical constraints in the experimental parameter semantic knowledge base. Credibility is a quantitative score obtained by weighted comprehensive evaluation of unit consistency, parameter logical relationships, and the reasonableness of the numerical range.

[0038] Consistency verification of standardized data is performed based on parameter association constraints in the experimental parameter semantic knowledge base. First, unit consistency verification is performed, converting the same physical quantity output from different devices in different units to a standardized unit system. The unit conversion verification is represented as follows: ; in, These are the original values. For unit conversion factor, The values ​​are in the standard unit system. Convert the offset to units.

[0039] Next, parameter logic relationships are verified to confirm whether the parameters meet the preset physical constraints, such as the compensation relationship between pH value and temperature, and the correlation between mass and calibration status. Simultaneously, abnormal data detection is performed, identifying outliers outside the reasonable range based on statistical methods and physical constraints. For example, if a pH value obtained from a certain analysis is 12.5, while the reasonable range for the device's historical data is 4.0 to 9.0, it is determined to be an outlier.

[0040] Based on this, a credibility assessment model is constructed, which performs a weighted comprehensive evaluation of unit consistency, parameter logical relationships, and the reasonableness of numerical range. The credibility assessment model is expressed as: ; in, Score the reliability of the analysis results. Scoring for consistency among units. Scoring based on the logical relationship of the parameters. Score the reasonableness of the numerical range. , , The balance coefficients are set to 0.3, 0.4, and 0.3 respectively, satisfying the following conditions: ,and For analysis results with a reliability score below 0.5, they are marked as abnormal data and subject to manual review. This ensures the accuracy and reliability of the analysis results.

[0041] Step S6: Output standardized experimental data that has been verified and evaluated for reliability. Standardized experimental data should include at least the device's unique identifier, experimental timestamp, parameter name, standard value, standard unit, and reliability score. It should be output in JSON or CSV format and stored in the experimental database.

[0042] Standardized experimental data, verified and assessed for reliability, is output in a unified format. A JSON or CSV data file is generated, with the unified data format including a unique device identifier, experimental timestamp, parameter name, standard value, standard unit, and reliability score. An experimental database is established to support the retrieval, comparison, and trend analysis of historical experimental data. The final output is a complete parsing report containing device information, parsed experimental data, and a reliability score, enabling intelligent acquisition, automatic parsing, semantic fusion, and reliable management of data from non-networked heterogeneous experimental devices. In one embodiment, such as... Figure 6 The schematic diagram of the data semantic perception and adaptive parsing system for non-networked devices in this embodiment is shown. A knowledge base of device resources and parsing rules is formed using heterogeneous non-networked experimental devices. Device semantic perception and parsing are then performed, followed by data storage. The non-networked device parsing model includes a device identification perception layer and a parsing output layer. These layers perform semantic perception recognition and parameter matching on the data to complete device identification. After parsing, a unified format is output, and parameter verification and feasibility assessment are conducted.

[0043] In one specific embodiment, to verify the effectiveness of this technical solution, an experiment was conducted on a simulation platform. The simulation conditions and parameters were set as follows: the experiment was conducted on an existing Python 3.13 and PyTorch 2.0.0 platform, using 2000 sets of experimental data collected from 10 types of laboratory equipment, including 1400 sets for training, 300 sets for validation, and 300 sets for testing. The learning rate was set to 0.0001, and the batch size was set to 8. The feature vector dimension of the device feature recognition module was set to 128 dimensions, the embedding vector dimension of the semantic annotation module was set to 256 dimensions, and the nearest neighbor number for transfer learning similarity matching was set to 3. Under the above conditions, the results were verified from four aspects: parsing accuracy, parameter semantic recognition accuracy, parsing rule configuration time, and credibility evaluation.

[0044] Analysis accuracy: Reference Figure 2 This paper demonstrates a comparison of the analytical accuracy of this invention with other analytical methods across different device types. The pH meter and digital vernier caliper were set as unknown devices during the training phase to verify transfer learning capabilities. Figure 2 As shown, this embodiment outperforms the comparison method on all types of devices. The overall parsing accuracy reaches over 85% on known devices, and still achieves approximately 80% parsing accuracy on unknown devices through transfer learning. In contrast, the accuracy of the comparison method drops significantly by 30% to 45% on unknown devices, verifying the effectiveness and generalization ability of the adaptive parsing module.

[0045] Parameter semantic recognition accuracy: Reference Figure 3 This paper demonstrates a comparison of the parametric semantic recognition accuracy of this invention with other semantic recognition methods under different data acquisition conditions. For example... Figure 3 The embodiment shown maintains a high recognition accuracy under all acquisition conditions, reaching 98% under standard format output conditions and 88.5% under screen OCR recognition conditions, while the comparison method only achieves about 50%~55% under OCR conditions. Compared with the regular expression matching method, the average accuracy is improved by about 20 percentage points under five conditions, verifying the robustness of the semantic perception module under different data quality conditions.

[0046] Parsing rule configuration time: Reference Figure 4 This paper compares the rule configuration time of this invention with other configuration methods under different numbers of new device connections. When connecting the first device, it takes approximately 18 minutes to establish the initial template, slightly longer than the rule base matching method but significantly shorter than the manual configuration method. From the second device onwards, the advantages of transfer learning gradually become apparent, requiring only about 10 minutes per new device. The total time for the 10th device is 134 minutes, saving approximately 84.5% compared to the manual configuration method and approximately 51.9% compared to the rule base matching method, verifying its significant advantage in reducing the cost of connecting new devices.

[0047] Credibility assessment: final Figure 5 This paper demonstrates the standardized data output and reliability assessment results of this invention on 20 representative test samples. Reliability levels are categorized as high reliability ≥ 0.8, medium reliability 0.5~0.8, and low reliability ≤ 0.5, requiring manual verification. All samples have a reliability higher than 0.75, with most reaching above 0.90. High-reliability samples account for 60%, and there are no low-reliability samples. For input data with inconsistent units, such as the Fahrenheit output "381.2F" from an eight-basket oven, it is automatically converted to the standard unit °C, achieving a reliability of 0.92. For input data containing OCR recognition errors, such as the pH meter output corrected to 9.96, the reliability is 0.82. For multi-parameter mixed data, it can correctly identify each parameter simultaneously. This verifies that the semantic consistency, unit uniformity, and reliability of the data are guaranteed while completing data parsing.

[0048] The simulation experiments described above demonstrate that this technology has significant advantages over existing technologies in terms of parsing accuracy, semantic recognition accuracy, configuration efficiency, and data reliability. It can effectively solve the technical problems of automatic data acquisition, semantic-level understanding, and adaptive parsing of data from non-networked heterogeneous experimental devices.

[0049] In this embodiment, a data semantic perception and adaptive parsing method for non-networked heterogeneous experimental devices is proposed. On the one hand, the device knowledge base and transfer learning mechanism are used to automatically identify device types and dynamically generate parsing rules, enabling the system to automatically adapt the parsing strategy when accessing unknown devices, effectively reducing the manual configuration cost of new device access. On the other hand, semantic annotation and semantic mapping are used to uniformly express the heterogeneous parameters output by different devices into a standard semantic space, realizing the leap from field-level parsing to semantic-level understanding of data. This effectively solves the problems of existing technologies lacking semantic understanding capabilities and being unable to uniformly identify and align heterogeneous parameters, and can maintain a high semantic recognition accuracy under diverse data formats.

[0050] Example 2 This invention also proposes a data semantic awareness and adaptive parsing system for non-networked devices, using the data semantic awareness and adaptive parsing method for non-networked devices as described in Example 1, and Figure 7 The system diagram shows the following features: The equipment semantic knowledge base construction module is used to collect basic information of experimental equipment, establish equipment category tags, experimental parameter semantic knowledge bases, and equipment parsing rule bases. Among them, the experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information, and the correlation constraints between parameters, while the equipment parsing rule base includes data format templates, field separators, and numerical extraction rules for various types of equipment.

[0051] The experimental data acquisition module automatically identifies the interface type of experimental devices based on device category tags and a device resolution rule base. It then acquires multi-source experimental data from non-networked experimental devices and caches it uniformly. Interface types include serial port, USB, Ethernet, and image acquisition interfaces. For devices supporting serial or USB communication, identification and connection are achieved by reading device descriptors and communication parameters. For devices that only display data on a screen, screen images are captured via a camera, and data is extracted using OCR technology.

[0052] The data semantic awareness module is used to semantically annotate cached experimental data based on an experimental parameter semantic knowledge base, identify experimental parameters, units, and dimensions, and map the identified parameters to a standard semantic space to obtain pre-parsing data with semantic labels. It uses a neural network to semantically annotate experimental data, calculating the semantic similarity between parameters and standard parameters to achieve the mapping of parameters to the standard semantic space.

[0053] The adaptive parsing module is used to perform structured parsing of the experimental data by calling the corresponding parsing rules in the device parsing rule library based on the semantic tags and device features in the data before parsing, resulting in standardized data. For unknown devices not recorded in the device parsing rule library, the module transfers parsing rules based on the feature similarity between the unknown device and known devices and adaptively adjusts them to generate parsing rules adapted to the unknown device. The transferred parsing rules are selected as the initial templates by choosing the parsing rule most similar to the known device, and then fine-tuned according to the actual output format of the unknown device.

[0054] The knowledge-driven verification module performs consistency verification on standardized data based on parameter association constraints in the experimental parameter semantic knowledge base. It evaluates the credibility of the parsing results based on the verification results and marks parsing results with credibility below a preset threshold as requiring further review. Consistency verification includes unit consistency verification and parameter logical relationship verification. Credibility is determined by a weighted comprehensive evaluation of unit consistency, parameter logical relationships, and the reasonableness of the numerical range.

[0055] The standardized data output module is used to output standardized experimental data that has been verified and assessed for reliability. The data output by the standardized data output module should include at least the device identifier, timestamp, parameter name, standard value, standard unit, and reliability score.

[0056] This embodiment constructs a data credibility verification method constrained by semantic knowledge. It incorporates experimental domain knowledge, physical constraints of parameters, and experimental process logic into the data parsing process. Multi-level consistency verification of the parsing results is performed, and a credibility assessment model is established. This enables credibility assessment and automatic error correction of experimental data, effectively solving the technical problems of insufficient credibility guarantee in parsing results and difficulty in automatically identifying and correcting abnormal data in existing technologies. It can also automatically identify and correct various data anomalies such as inconsistent units and OCR recognition errors. While completing data parsing, it ensures semantic consistency, unit uniformity, and credibility of the data, providing a reliable data foundation for subsequent intelligent analysis of experimental data.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A data semantic perception and adaptive parsing method for non-networked devices, characterized in that, Includes the following steps: Step S1: Construct a semantic knowledge base for non-networked devices, collect basic information about experimental devices, and establish a semantic knowledge base for device category tags, experimental parameters, and device parsing rules. Step S2: Based on the device category label and the device parsing rule base, automatically identify the interface type of the experimental device, and obtain multi-source experimental data from non-networked devices and cache them uniformly based on the identification results; Step S3: Based on the experimental parameter semantic knowledge base, perform semantic annotation on the cached experimental data, identify experimental parameters, units and dimensions, and map the identified parameters to the standard semantic space to obtain the pre-parsing data with semantic tags. Step S4: Based on the semantic tags and device features in the pre-parsing data, call the corresponding parsing rules in the device parsing rule base to perform structured parsing on the experimental data to obtain standardized data; for unknown devices not recorded in the device parsing rule base, transfer the parsing rules based on the feature similarity between the unknown device and known devices and adaptively adjust them to generate parsing rules adapted to the unknown device; Step S5: Based on the parameter association constraint relationship in the experimental parameter semantic knowledge base, perform consistency verification on the standardized data, evaluate the credibility of the parsing results based on the verification results, and mark the parsing results with credibility lower than the preset threshold as pending review; Step S6: Output standardized experimental data that has been verified and evaluated for reliability.

2. The data semantic perception and adaptive parsing method for non-networked devices according to claim 1, characterized in that, In step S1, the experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information, and the correlation constraints between parameters; the equipment parsing rule base includes data format templates, field separators, and numerical extraction rules for various types of equipment; the equipment category labels include at least pH meters, digital vernier calipers, electronic balances, and chromatographs.

3. The data semantic perception and adaptive parsing method for non-networked devices according to claim 2, characterized in that, In step S2, the interface types include serial port, USB, Ethernet, and image acquisition interface; for devices that support serial port or USB communication, identification and connection are achieved by reading the device descriptor and communication parameters; for devices that only display data on the screen, screen images are captured by a camera and data is extracted using OCR technology; the unified cache includes device number, acquisition time, original data fields, and data length information.

4. The data semantic perception and adaptive parsing method for non-networked devices according to claim 1, characterized in that, In step S3, the semantic annotation is implemented based on a neural network model, and the annotation results include parameter category labels, unit labels, and experimental condition labels. The mapping to the standard semantic space is achieved by calculating the semantic similarity between the target device parameters and the standard parameters in the experimental parameter semantic knowledge base. Parsing results with a credibility lower than a preset threshold are marked as pending review.

5. The data semantic perception and adaptive parsing method for non-networked devices according to claim 1, characterized in that, In step S4, the device features are based on semantic tags and device communication features in the pre-parsing data; the parsing rules include at least a data format template, field separators, numerical extraction positions, and unit conversion coefficients; the migration parsing rules are as follows: calculate the feature similarity between the unknown device and each known device in the device parsing rule library, select the parsing rule of the most similar known device as the initial template, and fine-tune the field separators and numerical extraction positions in the initial template according to the actual output format of the unknown device.

6. The data semantic perception and adaptive parsing method for non-networked devices according to claim 5, characterized in that, In step S5, the consistency verification includes unit consistency verification and parameter logical relationship verification; the unit consistency verification is to uniformly convert the same physical quantity in different unit systems to the standard unit system. The parameter logical relationship verification is to verify whether the parameters in the parsing result satisfy the preset physical constraint relationship in the experimental parameter semantic knowledge base; the credibility is a quantitative score obtained by weighted comprehensive evaluation of unit consistency, parameter logical relationship and numerical range rationality.

7. The data semantic perception and adaptive parsing method for non-networked devices according to claim 6, characterized in that, In step S6, the standardized experimental data includes at least the unique device identifier, experimental timestamp, parameter name, standard value, standard unit, and reliability score, and is output in JSON or CSV format and stored in the experimental database.

8. A data semantic perception and adaptive parsing system for non-networked devices, characterized in that, The system is implemented based on the method of any one of claims 1 to 7, comprising: The equipment semantic knowledge base construction module is used to collect basic information about experimental equipment and establish equipment category tags, experimental parameter semantic knowledge bases, and equipment parsing rule bases. The experimental data acquisition module is used to automatically identify the interface type of the experimental equipment according to the equipment category label and the equipment parsing rule base, and to obtain multi-source experimental data from non-networked experimental equipment and cache it uniformly according to the identification results; The data semantic awareness module is used to perform semantic annotation on the cached experimental data based on the experimental parameter semantic knowledge base, identify experimental parameters, units and dimensions, map the identified parameters to the standard semantic space, and obtain the parsed data with semantic tags. The adaptive parsing module is used to perform structured parsing of the experimental data by calling the corresponding parsing rules in the device parsing rule library based on the semantic tags and device features in the data before parsing, so as to obtain standardized data; for unknown devices not recorded in the device parsing rule library, the parsing rules are transferred and adaptively adjusted based on the feature similarity between the unknown device and the known device to generate parsing rules adapted to the unknown device. The knowledge-driven verification module is used to perform consistency verification on the standardized data based on the parameter association constraint relationship in the experimental parameter semantic knowledge base, evaluate the credibility of the parsing results based on the verification results, and mark the parsing results with credibility lower than a preset threshold as pending verification. The standardized data output module is used to output standardized experimental data that has been verified and evaluated for reliability.

9. The data semantic perception and adaptive parsing system for non-networked devices according to claim 8, characterized in that, In the device semantic knowledge base construction module, the experimental parameter semantic knowledge base includes parameter names, physical units, dimensional information and the correlation constraints between parameters, and the device parsing rule base includes data format templates, field separators and numerical extraction rules for various types of devices. In the experimental data acquisition module, the interface types include serial port, USB, Ethernet and image acquisition interface. For devices that support serial port or USB communication, identification and connection are achieved by reading device descriptors and communication parameters. For devices that only display data on the screen, screen images are acquired by a camera and data is extracted using OCR technology.

10. The data semantic perception and adaptive parsing system for non-networked devices according to claim 8, characterized in that, The data semantic perception module performs semantic annotation on experimental data based on neural networks, and realizes the mapping of parameters to the standard semantic space by calculating the semantic similarity between the parameters and standard parameters; the migration parsing rule in the adaptive parsing module is to select the parsing rule of the most similar known device as the initial template, and then fine-tune it according to the actual output format of the unknown device. The consistency verification in the knowledge-driven verification module includes unit consistency verification and parameter logical relationship verification. The credibility is a weighted comprehensive evaluation of unit consistency, parameter logical relationship and the rationality of numerical range. The data output by the standardized data output module includes at least device identifier, timestamp, parameter name, standard value, standard unit and credibility score.

Images