Intelligent field identification method and system for heterogeneous system

Abstract

The embodiment of the invention relates to the technical field of information, in particular to an intelligent field identification method and system oriented to a heterogeneous system. The method comprises the steps of collecting a DOM structure and context text information of a document object model of a target webpage; dOM elements are extracted, and DOM features are generated by combining context text information of the DOM elements and by means of a semantic model which is trained in advance and fused into a basic governance term library; screening out DOM (Document Object Model) features containing fields to be identified to form a library to be identified; calculating the similarity between the DOM features and element features in a preset element feature library; according to the similarity and a preset priority rule, determining element features matched with the to-be-identified field, and obtaining a matched field according to the element features; monitoring in real time and updating the to-be-recognized library according to changes of webpage elements; the DOM features comprise ID features, tag features, context features and visual features.

Description

Technical Field

[0001] The present application relates to the field of information technology, and in particular to an intelligent field recognition method and system for heterogeneous systems. Background Art

[0002] Field matching currently accounts for a significant portion of the daily maintenance workload in some information systems. Field matching is an unavoidable and urgent task, directly impacting the normal operation of other system functions. Field recognition in heterogeneous systems primarily relies on precise matching based on IDs / names, rule-driven pattern matching, and manual annotation mapping. Rule-driven approaches utilize regular expressions or templates to match field labels. However, these technologies have significant limitations and insufficient adaptability. They also lack the ability to understand and match unstructured or semantically similar fields, such as "household registration address" and "residential address." Furthermore, traditional methods are ineffective in complex scenarios like dynamic loading and nested forms, resulting in an adaptation rate of less than 25% in grassroots temporary reporting systems. Furthermore, the lack of industry terminology optimization in general-purpose tools results in field recognition accuracy exceeding 20% below actual requirements. Therefore, a new intelligent field recognition method is urgently needed to address these issues and improve cross-system adaptability and field recognition accuracy. Summary of the Invention

[0003] Multiple embodiments of this specification describe an intelligent field recognition method and system for heterogeneous systems.

[0004] In a first aspect, the embodiments of this specification provide a method for intelligent field identification for heterogeneous systems, including the following steps:

[0005] Collect the DOM structure and contextual text information of the document object model of the target web page;

[0006] Extract DOM elements, combine their contextual text information, and use a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generate DOM features based on the semantic description;

[0007] Filter out the DOM features containing the fields to be identified from all DOM features to form a library to be identified;

[0008] Calculating the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library;

[0009] Determine, based on the similarity and a preset priority rule, element features that match the field to be identified, and obtain matching fields based on the element features;

[0010] Monitor in real time and update the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages;

[0011] The DOM features include ID features, tag features, context features, and visual features, wherein the priority rule at least includes matching in the order of ID features, tag features, context features, and visual features.

[0012] In a second aspect, the embodiments of this specification provide an intelligent field recognition system for heterogeneous systems, including:

[0013] The acquisition module acquires the DOM structure and contextual text information of the document object model of the target web page;

[0014] An extraction module extracts DOM elements, combines their contextual text information, and uses a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generates DOM features based on the semantic description;

[0015] The screening module selects the DOM features containing the fields to be identified from all DOM features to form a library to be identified;

[0016] A calculation module calculates the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library;

[0017] A matching module, which determines element features that match the field to be identified based on the similarity and a preset priority rule, and obtains matching fields based on the element features;

[0018] A monitoring module monitors in real time and updates the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages;

[0019] The DOM features include ID features, tag features, context features, and visual features, wherein the priority rule at least includes matching in the order of ID features, tag features, context features, and visual features.

[0020] In a third aspect, embodiments of this specification provide an electronic device, including a processor and a memory;

[0021] The processor is connected to the memory;

[0022] The memory is used to store executable program code;

[0023] The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method described in any one of the above aspects.

[0024] In a fourth aspect, an embodiment of this specification provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in any one of the above aspects is implemented.

[0025] In a fifth aspect, embodiments of this specification provide a computer program product, including a computer program, which implements the method described in any of the above aspects when executed by a processor.

[0026] The beneficial effects of the technical solutions provided by some embodiments of this specification include at least:

[0027] In multiple embodiments of this specification, the provided intelligent field recognition method and system achieve a more comprehensive semantic understanding of DOM elements by combining ID features, tag features, context features, and visual features, and assigning different weights for weighted calculation, thereby improving the accuracy of field matching. It reduces the workload of manually configuring static mapping tables and reduces the complexity of operation and maintenance, and is particularly suitable for data source systems that need to be frequently updated or upgraded. When the meaning of a field changes or new field requirements are introduced, the field mapping strategy can be quickly followed up and adjusted. It monitors changes in web page elements to ensure that dynamically loaded new form items can be automatically detected and processed, expanding the scope of application.

[0028] Other features and advantages of the various embodiments of this specification will be further disclosed in the following detailed description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0030] Figure 1 This is a schematic diagram of the intelligent field recognition method and system application provided in the embodiments of this specification.

[0031] Figure 2 This is a flow chart of the intelligent field recognition method provided in the embodiments of this specification.

[0032] Figure 3 This is a schematic diagram of DOM elements provided in the embodiments of this specification.

[0033] Figure 4 Schematic diagram of the intelligent field recognition system provided in the embodiments of this specification.

[0034] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0035] The following is an explanation and description of the technical solutions of the embodiments of this specification in conjunction with the drawings of the embodiments of this specification. However, the following embodiments are only preferred embodiments of this specification and are not exhaustive. Based on the embodiments in the implementation mode, other embodiments obtained by those skilled in the art without making any creative work are all within the scope of protection of this specification.

[0036] Throughout this specification, the claims, and the accompanying drawings, the terms "first," "second," "third," and the like are used to distinguish between different items, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may include other steps or elements inherent to the process, method, product, or apparatus.

[0037] In the following description, terms such as "inside", "outside", "up", "down", "left", "right", etc. that indicate directions or positional relationships are only used to facilitate the description of the embodiments and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, they should not be understood as limitations on this specification.

[0038] The data involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions.

[0039] Before introducing the technical solution in this specification, the application scenarios and related technologies of the technical solution are introduced.

[0040] Field mapping refers to establishing a correspondence between fields between different data sources or systems in order to achieve data exchange, conversion and integration. It is used to identify the same or similar data items in different systems and associate them so that data can be smoothly migrated from one system to another, or processed uniformly in an integrated environment. Field mapping can be done by manually configuring a static mapping table, or by using automated tools such as ETL (Extract, Transform, Load) software to achieve dynamic mapping. For example, the grassroots governance system covers a wide range of public service functions, including population management, social security services, medical insurance, education support, etc. It is necessary to process data from different departments and sources, so effective field mapping is crucial to ensure data accuracy and consistency. For example, see the attached Figure 1During the low-income benefit application process, key fields such as "household registration address," "income status," and "family member information" must be accurately matched to ensure the proper protection of individual rights. The system is required to not only recognize these fields but also perform appropriate conversion and adaptation based on specific application scenarios.

[0041] During the deployment phase of a new system, a detailed analysis of each participating data source is required to determine which fields are required and how to match them with the corresponding fields in the target system. Over time, the source system may undergo updates or upgrades, resulting in changes in field names or formats. At this time, the field mapping rules need to be adjusted to maintain the consistency of the data flow. The operation and maintenance team is required to follow up and adjust the field mapping strategy in a timely manner. For example, whenever a new policy is implemented, additional information may need to be collected. Sometimes even if the field name remains unchanged, its meaning may change with policy adjustments. Compatibility issues, due to the possibility of different implementation details in different regions, cross-regional data integration becomes more complicated. For example, the criteria for identifying "low-income people" may vary from place to place, which requires customized adjustment of the field mapping solution for specific regions.

[0042] The use of intelligent field recognition methods for heterogeneous systems can significantly reduce the burden of operation and maintenance. Figure 2 This specification provides an intelligent field recognition method for heterogeneous systems, including the following steps:

[0043] Step S1) collecting the DOM structure and contextual text information of the document object model of the target web page 10.

[0044] DOM, or Document Object Model, is a web page standard developed by the W3C. It provides common methods and properties for HTML and XML documents. Use a browser automation tool (for example, Selenium WebDriver, Playwright, or Puppeteer) to load the target web page.10 Ensure that the web page is fully loaded, including any asynchronously loaded content (such as AJAX requests). You can confirm that the page has loaded by listening for specific events (such as DOMContentLoaded or load) or by waiting for a specific element to appear. Use JavaScript to access the DOM structure of the web page. You can access the HTML content of the entire page using document.documentElement.outerHTML. For each DOM element, extract the surrounding contextual text information. By traversing the DOM tree, find labels or other descriptive text near each input field. Organize the collected information into a structured format.

[0045] Please see the attached Figure 3 The DOM features include ID features 12, tag features 13, context features 14, and visual features 15, wherein the priority rule at least includes matching in the order of ID features 12, tag features 13, context features 14, and visual features 15. ID feature 12 is "householdRegistrationGroup". Tag features 13 include:

[0046] “

[0047] <label for="householdRegistrationAddress"> New address:< / label>

[0048] <input type="text" id="householdRegistrationAddress" name="householdRegistrationAddress" placeholder="请输入您的新住址">

[0049] ”.

[0050] The contextual features 14 include the content of other parts of the web page, while the visual features 15 are determined by the CSS code and are visible to the page viewer.

[0051] Step S2) extracts DOM elements, combines them with contextual text information, and uses a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generates DOM features based on the semantic description.

[0052] Specifically, it includes: standardizing DOM elements and obtaining the hierarchical depth of DOM elements in the document;

[0053] Obtain contextual text information corresponding to the DOM element according to the hierarchical depth;

[0054] Based on the pre-trained semantic model integrated into the grassroots governance terminology library, the standardized DOM elements and their corresponding contextual text information are semantically analyzed to obtain semantic descriptions;

[0055] Using a semantic vector to represent the semantic description, adding the visual feature 15 to the semantic vector;

[0056] Extracting feature vectors corresponding to the ID feature 12, the tag feature 13, the context feature 14, and the visual feature 15 from the semantic vector, and splicing all the feature vectors in a predetermined splicing order as DOM features;

[0057] According to the tag feature 13 , DOM features including the to-be-identified field 11 are screened out from all DOM features to form a to-be-identified library.

[0058] The standardization process includes unifying the formats of IDs and tags, converting all ID and tag names to lowercase, and removing special characters.

[0059] The contextual text information obtained includes two types: one is the key information of the page, including the website name, page name, several page titles, and the overall page summary. The other type is the names and labels of all parent nodes of the DOM tree where the DOM element resides, as well as the names and labels of all sibling nodes. Contextual text information helps understand the position of the DOM element within its document structure and its surrounding environment. This information is very useful for tasks such as field identification and data extraction, especially when automating complex web page structures.

[0060] The input of the semantic model is the DOM element and its corresponding contextual text information, for example, as shown in the following Figure 1 The DOM element in the text is "address," and its context includes the webpage title "Floating Population Registration" and "residential address." After training, the "Floating Population Registration" semantic model can generate a semantic description: Based on the DOM element "address" and its context (the page title "Floating Population Registration" and the adjacent text "residential address"), it is determined that this field is used for floating population information registration. The DOM element "address" in this page is the household registration address, which is represented by address@hukouben.

[0061] Encode text information (such as tags, context, and field names) into a high-dimensional numerical vector, where each dimension represents a semantic feature. The DOM element's ID feature 12 is "householdRegistrationGroup," the tag feature 13 is "address," and the context feature 14 is "page title ("Floating Population Registration")," and the adjacent tag is "address." When extracting visual feature 15, the element's position, size, font, font size, and type are extracted from the DOM element. For example, the vector representing visual feature 15 is: [120, 450, 300, 30,"Arial,"14px,"text"], where x=120, y=450, width=300, height=30, font "Arial," font size "14px," and type "text." By assigning numbers to "Arial," "14px," and "text," the vector representing visual feature 15 can be represented as: [120, 450, 300, 30, 1, 2, 1]. Accordingly, ID feature 12: householdRegistrationGroup, tag feature 13: address, context feature 14: page title "Floating Population Registration", and adjacent tag "address" are all limited to a certain length. A fixed-length number is used to represent the text. If the text is not long enough, a preset empty value is used instead. The final DOM feature is:

[0062] The semantic vector and the visual feature 15 are concatenated into a complete feature vector as the semantic vector of the field. The feature vectors corresponding to the ID feature 12, the tag feature 13, the context feature 14, and the visual feature 15 are extracted and all the feature vectors are concatenated in a predetermined concatenation order as the DOM feature.

[0063] Exemplarily, string encoding rules include using UTF-8 character encoding for all text, with each character represented by its corresponding ASCII or Unicode code. Fixed length restrictions include: ID feature 12 can be a maximum of 25 characters, tag feature 13 can be a maximum of 10 characters, and context feature 14 can be a maximum of 30 characters (the concatenation of the page title and adjacent tags). Insufficient lengths are padded with preset null values (e.g., 0). For example, if ID feature 12 is "householdRegistrationGroup" (actual length > 25), any excess characters are truncated, resulting in "householdRegistrationGrou"). Tag feature 13 is "address," and context feature 14 is "the webpage title 'Migrant Population Registration', and the adjacent tag 'Address'." After corresponding conversion into numerical values representing the text, a feature vector is obtained, which is then arranged in a predetermined order to obtain a DOM feature. For example, the DOM features are: [#ID feature 12 (25 dimensions) 104, 111, 117, 115, 101, 104, 116, 111, 450, 300, 30, 108, 103, 100, 82, 101, 103, 105, 115, 124, 182, 117, 116, 101, 23, #Tag feature 13 (10 dimensions) 22320, 22328,0,0,0,0,0,0,0,0,#Contextual features 14 (30 dimensions) 27969,21160,20154,21475,21488,35760,20837,32467,22343,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,#Visual features 15 (7 dimensions) 120,450,300,30,1,2,1].

[0064] As an example, the method of training a semantic model includes:

[0065] Collect and organize professional terms and common vocabulary in the field of grassroots governance and build a grassroots governance terminology database;

[0066] Generate a BERT model and pre-train it on a general-purpose large-scale corpus to gain broad semantic understanding capabilities;

[0067] Based on pre-training, the BERT model is fine-tuned using the grassroots governance terminology library;

[0068] Receive and use the labeled dataset to test and re-fine-tune the fine-tuned model;

[0069] Based on the tested BERT model, a semantic model is obtained.

[0070] Collect and organize documents related to grassroots governance, governance system field definitions, commonly used business forms, and frequently used vocabulary used by the public to build a professional terminology library covering core grassroots governance concepts. This terminology library is used to enhance the semantic model's understanding within specific domains and improve the accuracy of field recognition and matching. Example terms include, but are not limited to: population management terms such as household registration address, migrant population, and family member relationships; social security terms such as pension, medical insurance enrollment, and benefit level; and form field terms such as address, contact number, income, and workplace.

[0071] An initial language model was built based on the open-source BERT architecture (such as BERT-base or RoBERTa) and pre-trained using a large-scale general Chinese corpus (such as Chinese Wikipedia, Baidu Baike, and news corpora). Based on this general pre-training, the BERT model was further fine-tuned for domain adaptability using the constructed grassroots governance terminology library. Specifically, this includes:

[0072] Inject a term dictionary into the model vocabulary as additional input;

[0073] Continue training using domain text data containing term context;

[0074] The model output layer is adjusted to better capture the semantic associations between governance fields.

[0075] Through fine-tuning at this stage, the model can accurately identify the meaning of contextually relevant fields such as "household registration address" and "whether a household registration has been established." Subsequently, field samples from real-world governance systems were collected and combined with manually annotated data (such as field labels, contextual information, and DOM structure) to form a high-quality annotated dataset. This dataset was used to evaluate the performance of the fine-tuned BERT model. Based on the evaluation results, a second round of fine-tuning and optimization was performed to further improve the model's generalization and recognition accuracy in real-world application scenarios. After this training and optimization, a semantic model was obtained that integrates general semantic understanding with domain knowledge of grassroots governance.

[0076] Step S3) DOM features including the to-be-identified field 11 are screened out from all DOM features to form a to-be-identified library.

[0077] For example, based on the tag feature 13, DOM features containing the to-be-identified field 11 are selected from all DOM features to form a to-be-identified library. If the tag feature 13 is "address," since the meaning of "address" on this page is broad and has multiple ambiguous interpretations, it will be included in the to-be-identified library as the to-be-identified field 11.

[0078] Step S4) Calculate the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library. Specifically including:

[0079] Obtaining semantic vectors of element features in the element feature library;

[0080] For each field to be identified 11, obtain the DOM feature where it is located;

[0081] For element features in the element feature library, the vector similarity between the DOM feature and the semantic vector of the element feature is calculated.

[0082] For the target field to be identified on the current page, extract its corresponding DOM feature vector. This vector incorporates ID features 12, tag features 13, contextual features 14, and visual features 15 (such as position, size, and font). The DOM feature vector is ultimately a fixed-length numerical vector. For each candidate field (from the element feature library), calculate the vector similarity between its semantic vector and the current DOM feature vector. Common methods include cosine similarity, Euclidean distance, and Manhattan distance.

[0083] Methods for presetting element feature libraries include:

[0084] Read multiple sample web pages, extract DOM elements, and perform standardization on the DOM elements;

[0085] Combined with its contextual text information and with the help of a pre-trained semantic model integrated into the grassroots governance terminology library, a semantic description of the DOM element is generated;

[0086] Using a semantic vector to represent the semantic description, and generating a DOM feature according to the semantic vector;

[0087] The manually annotated fields of the DOM features are received, and an element feature library is generated according to the annotated DOM features.

[0088] Collect sample data from multiple governance system pages and extract key form field elements, including but not limited to: <input> Input box, <select> Drop-down selection box,< / select> <textarea> Multi-line text input box,< / textarea> <label>Label text, related container nodes (such as ). For each extracted DOM element, the following standardization operations are performed: unified ID format to remove special characters, convert to lowercase, unified tag names (such as input, select), hierarchical depth calculation to determine its nesting level in the DOM tree, and unified processing of attributes such as type, name, placeholder, etc.

[0089] For example, DOM elements <input id="HouseholdRegistrationAddress" type="text"> , after standardization, it becomes {"id": "householdregistrationaddress", "tag": "input", "depth":4}. For each DOM element, its contextual text information (such as label text, adjacent fields, form title, etc.) is obtained, and its corresponding semantic description is generated using a semantic model that integrates the grassroots governance terminology library.

[0090] Example of contextual information: Page title: "Migrant Population Registration," adjacent tags: "Address," "Contact Number." The semantic model generates the following result: "This field represents the household registration address, i.e., the legal residential address as recorded in the household register."

[0091] Using the trained BERT semantic model, the above semantic description is converted into a high-dimensional semantic vector, and combined with visual features15 (position, size, font, type, etc.) to generate a complete DOM feature vector.

[0092] Receive manual annotations and generate an element feature library. Submit the generated DOM feature vectors to the annotator, who will manually annotate them according to the actual field meaning. Please refer to Table 1 for an annotation example.

[0093] Table 1 Annotation examples

[0094]

[0095] The annotated field information is bound to its corresponding DOM feature vector to form a preset element feature library for subsequent field recognition, automatic mapping and other tasks.

[0096] Step S5) Determine the element features that match the to-be-identified field 11 based on the similarity and the preset priority rules, and obtain the matching field 21 based on the element features. Specifically, it includes:

[0097] Set weights for the matching degrees of the ID feature 12, the tag feature 13, the context feature 14, and the visual feature 15 respectively;

[0098] Try to match the DOM feature with the ID feature 12, tag feature 13, context feature 14, and visual feature 15 of the element feature using an exact matching method. If any feature matches, the preset maximum similarity score is used as the final similarity score.

[0099] If the exact match is unsuccessful, the similarity between the DOM feature and the ID feature 12, tag feature 13, context feature 14 and visual feature 15 of the element feature is calculated using a fuzzy matching method, and the weighted mean of the similarities is calculated using preset weights as the final similarity score;

[0100] According to the final similarity score, the element feature that matches the field to be identified 11 is determined, and the matching field 21 is obtained according to the element feature.

[0101] To improve matching accuracy, priority rules are introduced to weight or modify the scores based on the original semantic similarity. Each feature type is assigned a preset weight value, for example, ID feature 12 has a weight of 0.4, tag feature 13 has a weight of 0.3, context feature 14 has a weight of 0.2, and visual feature 15 has a weight of 0.1. Using an "exact match" approach, the current DOM feature is rigorously compared with the standard fields in the element feature library. Specifically, this involves checking whether ID feature 12 is completely consistent, tag feature 13 is completely matched, context feature 14 is highly correlated, and visual feature 15 meets the reasonableness requirements of the field type. If any key feature matches successfully (for example, ID or tag are completely consistent), the field is considered clearly identifiable, and the preset maximum similarity score (for example, 1.0) is directly used as the final similarity score for the field.

[0102] If no match is found through exact matching, "fuzzy matching" is used to continue processing. Similarity scores (such as cosine similarity, edit distance, keyword overlap rate, etc.) are calculated for the ID feature 12, tag feature 13, context feature 14, and visual feature 15 between the DOM feature and the candidate field.

[0103] Multiply the similarity scores of each feature by its corresponding weight and sum them to obtain the final weighted similarity score. If the highest score exceeds a preset threshold (e.g., 0.7), the match is considered successful; otherwise, it is marked as "unrecognized field," which may trigger manual review or prompt for additional annotation.

[0104] After assigning a unique number to the address of the DOM element in the page, a matching field 21 can be mapped to the DOM element in the field mapping table.

[0105] On the other hand, in another embodiment, the priority rules include: If the DOM tag is consistent with the element feature tag, the weight is increased. If it contains keywords such as "addr" and "household", the weight is increased. If the input box type is "text" and the size is reasonable, the weight is increased. If the context is highly relevant, the weight is appropriately increased. All candidate fields are sorted by the final similarity score, and the field with the highest score is selected as the matching result.

[0106] Step S6) monitors in real time and updates the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages.

[0107] In order to cope with the dynamic loading of content in modern web pages (such as form fields dynamically rendered through AJAX and JavaScript), this step detects new or changed field elements by monitoring the changes in the web page DOM structure and automatically adds them to the "to be identified library" to support the dynamic adaptation of the subsequent field identification process. Use the DOM observer (MutationObserver API) provided by the browser or the lifecycle hook provided by the front-end framework (such as useEffect of Vue / React) to monitor the DOM changes of key areas in the target web page 10 in real time. When a new field is detected, the same method as steps S1-S2 is followed. The newly extracted field feature vector is added to the "to be identified library" for subsequent field matching module calls. This method is applicable to the following scenarios: (1) the web page asynchronously loads new form items through JavaScript; (2) user interaction triggers the display of new fields (such as clicking the "Add Family Member" button); (3) the form dynamically displays different fields based on user input (such as clicking to select "Identity Type" to display the corresponding ID card field, passport field, residence permit field, etc.).

[0108] Whenever the library to be recognized is updated, the following processing options are available: automatically triggering the field recognition process (recommended), or waiting for an action (such as submitting a form) before performing recognition. At the same time, metadata such as the field source page URL and loading timestamp are recorded to facilitate subsequent log tracking and optimization analysis.

[0109] On the other hand, this specification provides an intelligent field recognition system for heterogeneous systems. Figure 4 ,include:

[0110] The acquisition module 100 acquires the DOM structure and contextual text information of the target web page 10;

[0111] Extraction module 200 extracts DOM elements, combines their contextual text information, and uses a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generates DOM features based on the semantic description;

[0112] The screening module 300 screens out DOM features containing the to-be-identified field 11 from all DOM features to form a to-be-identified library;

[0113] The calculation module 400 calculates the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library;

[0114] The matching module 500 determines the element features that match the to-be-identified field 11 according to the similarity and a preset priority rule, and obtains the matching field 21 according to the element features;

[0115] The monitoring module 600 monitors in real time and updates the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages;

[0116] The DOM features include ID features 12 , tag features 13 , context features 14 and visual features 15 , wherein the priority rule at least includes matching in the order of ID features 12 , tag features 13 , context features 14 and visual features 15 .

[0117] See also Figure 5 A schematic structural diagram of an electronic device provided in an embodiment of this specification is shown.

[0118] like Figure 5 As shown, the electronic device 1100 may include: at least one processor 1101, at least one network interface 1104, a user interface 1103, a memory 1105, and at least one communication bus 1102. The communication bus 1102 may be used to implement communication between the aforementioned components. The user interface 1103 may include buttons, and optionally may also include a standard wired interface or a wireless interface. The network interface 1104 may include, but is not limited to, a Bluetooth module, an NFC module, a Wi-Fi module, etc. The processor 1101 may include one or more processing cores. The processor 1101 utilizes various interfaces and circuits to connect the various components within the electronic device 1100. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 1105, and accessing data stored in the memory 1105, the processor 1101 performs various functions of the routing device and processes data. Optionally, the processor 1101 may be implemented in hardware using at least one of a DSP, an FPGA, and a PLA. The processor 1101 may integrate one or a combination of a CPU, a GPU, and a modem. Among them, the CPU mainly processes the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content that needs to be displayed on the display; and the modem is used to handle wireless communication.

[0119] It is understandable that the above-mentioned modem may not be integrated into the processor 1101, but may be implemented by a separate chip.

[0120] Memory 1105 may include either RAM or ROM. Optionally, memory 1105 may include non-transitory computer-readable media. Memory 1105 may be used to store instructions, programs, codes, code sets, or instruction sets. Memory 1105 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, sound playback function, image playback function, etc.), instructions for implementing the aforementioned method embodiments, etc.; the data storage area may store data related to the aforementioned method embodiments, etc. Memory 1105 may also optionally be at least one storage device located remotely from the aforementioned processor 1101. Memory 1105, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs. Processor 1101 may be configured to invoke the application programs stored in memory 1105 and execute the methods described in the aforementioned embodiments.

[0121] The embodiments of this specification also provide a computer-readable storage medium having instructions stored therein that, when executed on a computer or processor, cause the computer or processor to perform the steps of the aforementioned embodiments. If the components of the aforementioned electronic device are implemented as software functional units and sold or used as independent products, they may be stored in the computer-readable storage medium.

[0122] The embodiments of this specification also provide a computer program product, including a computer program, which implements multiple steps in the above embodiments when executed by a processor.

[0123] In the absence of conflict, the technical features in this embodiment and implementation scheme can be combined arbitrarily.

[0124] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product comprises multiple computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium accessible by a computer or a data storage device such as a server or data center that integrates multiple available media. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state drive (SSD)).

[0125] When implemented via hardware or firmware, the aforementioned method flow is programmed into the hardware circuit to create the corresponding hardware circuit structure and achieve the corresponding function. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit, whose logical function is determined by user programming of the device. Designers can "integrate" a digital system onto a PLD through their own programming, eliminating the need for chip manufacturers to design and fabricate dedicated integrated circuit chips. Furthermore, today, instead of manually fabricating integrated circuit chips, this programming is often performed using "logic compiler" software. This software is similar to the software compilers used in program development. However, the original code must be written in a specific programming language, called a hardware description language (HDL), before it can be compiled. There are many different types of HDL, and those skilled in the art will appreciate that simply programming the method flow using one of the aforementioned hardware description languages and then programming it into the integrated circuit can easily create a hardware circuit that implements the logic method flow.

[0126] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Without departing from the design spirit of this specification, various modifications and improvements made to the technical solutions of this specification by ordinary technicians in this field should fall within the scope of protection determined by the claims of this specification. < / label>

Claims

1. An intelligent field recognition method for heterogeneous systems, characterized in that: Including steps: Collect the DOM structure and contextual text information of the document object model of the target web page; Extract DOM elements, combine their contextual text information, and use a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generate DOM features based on the semantic description; Filter out the DOM features containing the fields to be identified from all DOM features to form a library to be identified; Calculating the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library; Determine, based on the similarity and a preset priority rule, element features that match the field to be identified, and obtain matching fields based on the element features; Monitor in real time and update the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages; The DOM features include ID features, tag features, context features, and visual features, wherein the priority rule at least includes matching in the order of ID features, tag features, context features, and visual features.

2. The intelligent field recognition method for heterogeneous systems according to claim 1, characterized in that: Extracting DOM elements, combining their contextual text information, and using a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generating DOM features based on the semantic description include: Standardize the DOM elements to obtain the hierarchical depth of the DOM elements in the document; the standardization includes unifying the formats of IDs and tags, Obtain contextual text information corresponding to the DOM element according to the hierarchical depth; Based on the pre-trained semantic model integrated into the grassroots governance terminology library, the standardized DOM elements and their corresponding contextual text information are semantically analyzed to obtain semantic descriptions; Using a semantic vector to represent the semantic description, and adding visual features to the semantic vector; Extracting feature vectors corresponding to ID features, tag features, context features, and visual features from the semantic vector, and splicing all feature vectors in a predetermined splicing order as DOM features; According to the tag features, DOM features containing the fields to be identified are screened out from all DOM features to form a library to be identified.

3. The intelligent field recognition method for heterogeneous systems according to claim 2, characterized in that: Methods for training semantic models include: Collect and organize professional terms and common vocabulary in the field of grassroots governance and build a grassroots governance terminology database; Generate a BERT model and pre-train it on a general-purpose large-scale corpus to gain broad semantic understanding capabilities; Based on pre-training, the BERT model is fine-tuned using the grassroots governance terminology library; Receive and use the labeled dataset to test and re-fine-tune the fine-tuned model; Based on the tested BERT model, a semantic model is obtained.

4. The intelligent field recognition method for heterogeneous systems according to claim 3 is characterized in that: The method for calculating the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library includes: Obtaining semantic vectors of element features in the element feature library; For each field to be identified, obtain its DOM feature; For element features in the element feature library, the vector similarity between the DOM feature and the semantic vector of the element feature is calculated.

5. The intelligent field recognition method for heterogeneous systems according to claim 4 is characterized in that: The method for determining the element features matching the to-be-identified field based on the similarity and the preset priority rules includes: Set weights for the matching degree of ID features, tag features, context features, and visual features respectively; Try to use an exact match method to match the DOM features with the ID features, tag features, context features, and visual features of the element features. If any feature matches, the preset maximum similarity score is used as the final similarity score; If the exact match is unsuccessful, the similarity between the DOM feature and the ID feature, tag feature, context feature, and visual feature of the element feature is calculated using a fuzzy matching method, and the weighted average of the similarities is calculated using preset weights as the final similarity score; According to the final similarity score, the element features that match the field to be identified are determined, and the matching fields are obtained according to the element features.

6. The intelligent field recognition method for heterogeneous systems according to any one of claims 1 to 4, characterized in that: Methods for presetting element feature libraries include: Read multiple sample web pages, extract DOM elements, and perform standardization on the DOM elements; Combined with its contextual text information and with the help of a pre-trained semantic model integrated into the grassroots governance terminology library, a semantic description of the DOM element is generated; Using a semantic vector to represent the semantic description, and generating a DOM feature according to the semantic vector; The manually annotated fields of the DOM features are received, and an element feature library is generated according to the annotated DOM features.

7. Intelligent field recognition system for heterogeneous systems, characterized by: include: The acquisition module acquires the DOM structure and contextual text information of the document object model of the target web page; An extraction module extracts DOM elements, combines their contextual text information, and uses a pre-trained semantic model integrated into the grassroots governance terminology library to generate a semantic description of the DOM elements, and generates DOM features based on the semantic description; The screening module selects the DOM features containing the fields to be identified from all DOM features to form a library to be identified; A calculation module calculates the similarity between the DOM features in the to-be-identified library and the element features in the preset element feature library; A matching module, which determines element features that match the field to be identified based on the similarity and a preset priority rule, and obtains matching fields based on the element features; A monitoring module monitors in real time and updates the to-be-identified library according to changes in web page elements to adapt to dynamically loaded web pages; The DOM features include ID features, tag features, context features, and visual features, wherein the priority rule at least includes matching in the order of ID features, tag features, context features, and visual features.

8. An electronic device, characterized in that: including a processor and a memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images