Website data collection strategy analysis system and method based on large language model

Abstract

The invention provides a website data collection strategy analysis system and method based on a large language model, and belongs to the technical field of website data collection strategy analysis. The method comprises the following steps: reading a target website domain name and predefined data collection terms, and identifying and crawling related page contents as candidate pages to extract a high-confidence data collection strategy page; then, a website data collection clause coverage condition analysis module based on a large language model analyzes data collection clauses followed and not followed by a target website by taking the segmented page fragments as input; the website personal data collection practice analysis module based on the large language model analyzes personal information types, collection intentions and third-party service types by taking the same fragment as input; and finally, integrating clause coverage and a personal data practice analysis result, and generating a final data collection strategy analysis report. The method has perfect understanding analysis capability and reasoning capability, and can realize accurate analysis and induction of website data collection strategy practice.

Description

Technical Field

[0001] The present invention relates to the technical field of website data collection strategy analysis, and in particular to a website data collection strategy analysis system and method based on a large language model. Background Art

[0002] Against the backdrop of the rapid development of informatization and digitalization, strengthening personal information protection and regulating the acquisition and use of personal information are crucial not only for the protection of individual rights and interests but also for the healthy development of the digital economy. To this end, the vast majority of online service websites are required to disclose their data collection practices through data collection policy documents, such as those discussing how online services collect user information and the rights users can exercise regarding their personal data. By analyzing website data collection policies, users can be provided with a clear and intuitive description of website data collection practices, thereby enhancing citizens' sense of digital security and promoting the healthy development of the digital economy.

[0003] However, given that data collection policy documents are often lengthy, complex, and formatted in varying ways, automated data collection policy analysis remains challenging. Existing techniques typically employ unsupervised, rule-based approaches or train and fine-tune machine learning classifiers, limiting their capabilities to analyzing a small set of manually predefined data collection clauses. Consequently, a comprehensive understanding and analysis of website data collection practices remains lacking.

[0004] Based on this, the public urgently needs an accurate and reliable automated website data collection strategy analysis method and system, which is of great significance to ensuring data security. Summary of the Invention

[0005] In view of the above background, the present invention aims to provide a website data collection strategy analysis system and method based on a large language model. The method and system have strong understanding and analysis capabilities and reasoning capabilities, and can realize efficient and accurate analysis of website data collection strategies.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a website data collection strategy analysis system based on a large language model, comprising:

[0008] An input module, used to read the target website domain name and predefined data collection terms of the website to be analyzed;

[0009] The data collection strategy crawling module is used to identify and crawl page contents related to the data collection strategy from the target website as candidate pages;

[0010] Data collection strategy preprocessing module, used to filter candidate pages and extract high-confidence data collection strategy pages;

[0011] The website data collection policy coverage analysis module, based on a large language model, takes contextual snippets from high-confidence data collection policy pages as input and uses the large language model to analyze the data collection policies that the target website complies with and does not comply with.

[0012] A website personal data collection practice analysis module based on a large language model uses contextual snippets extracted from high-confidence data collection strategy pages as input and employs a large language model to analyze the types of personal information collected by the target website, the intention behind collecting personal information, and the types of third-party services that receive personal information.

[0013] The output module is used to integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

[0014] As a preferred embodiment of the present invention, the predefined data collection terms for the website to be analyzed include the types of personal information collected, the types of personal information disclosed, the types of sources of personal information, the intention of collecting personal information, the intention of disclosing personal information, the types of third-party services that receive personal information, and the retention period of personal information.

[0015] As a preferred embodiment of the present invention, the data collection strategy crawling module is specifically configured as follows:

[0016] Extract keywords related to data collection terms from a pre-set multilingual vocabulary, and use keyword matching technology to identify hyperlinks on the target website's homepage that potentially point to the data collection policy page. The pages pointed to by the hyperlinks are considered candidate data collection policy pages.

[0017] Load the candidate data collection strategy page and download the HTML content. Remove the boilerplate content and convert it into a Markdown document. Each candidate data collection strategy page corresponds to a document.

[0018] As a preferred embodiment of the present invention, the data collection strategy preprocessing module is specifically configured as follows:

[0019] Apply pre-trained machine learning classifiers to initially screen candidate pages;

[0020] After the initial screening of the results through URL keyword filtering, the candidate pages that meet the URL keyword are retained; if the results after the initial screening do not meet the requirements, the URL keyword filtering process is skipped;

[0021] Match the target website domain name or redirected domain name based on the second-level domain name, exclude irrelevant pages, and use the remaining results as high-confidence data collection strategy pages;

[0022] Split the document corresponding to the high-confidence data collection strategy page into segments that meet the constraints of the large language model context window.

[0023] As a preferred embodiment of the present invention, the classification process of the pre-trained machine learning classifier is:

[0024] Taking the document corresponding to the high-confidence data collection strategy page as input, the feature vector is extracted based on the TextRank or PositionRank algorithm, and the feature vector is input into the random forest model to calculate the probability that the document content meets the data collection strategy page. The pages corresponding to the document content with a probability greater than the threshold are used as candidate pages after preliminary screening.

[0025] As a preferred embodiment of the present invention, the URL keywords are collected using data.

[0026] As a preferred embodiment of the present invention, the website data collection clause coverage analysis module based on the large language model is specifically configured as follows:

[0027] Instruct the large language model to play the role of an expert and provide a task description. The task description is to determine whether the input segment contains predefined information about the data collection terms of the website to be analyzed, and output the result in JSON format.

[0028] Integrate all the fragments of all high-confidence data collection strategy pages, take the union of the terms output by the large language model, and use it as the evaluation result of the website data collection terms coverage of the target website.

[0029] As a preferred embodiment of the present invention, the website personal data collection practice analysis module based on the large language model is specifically configured as follows:

[0030] Instruct the large language model to act as an expert and provide a task description. The task description is to classify the subcategories of personal data collection practices in the input segment and output a result in JSON format. The personal data collection practices include three categories: the type of personal information collected, the intention of collecting personal information, and the type of third-party services that receive personal information. Each category of personal data collection practices contains several subcategories.

[0031] Integrate all snippets of all high-confidence data collection strategy pages, take the union of each subcategory output by the large language model, and use it as the analysis and evaluation results of the target website's personal data collection practices.

[0032] In a second aspect, the present invention proposes a website data collection strategy analysis method based on the website data collection strategy analysis system, comprising the following steps:

[0033] (1) Define the terms of data collection for the website to be analyzed;

[0034] (2) Enter the target website domain name;

[0035] (3) Identify and crawl the page content related to the data collection strategy from the target website as candidate pages;

[0036] (4) Used to filter candidate pages and extract high-confidence data collection strategy pages;

[0037] (5) Using contextual snippets from high-confidence data collection policy pages as input, a large language model is used to analyze the data collection terms that the target website complies with and does not comply with;

[0038] (6) Using contextual fragments extracted from high-confidence data collection strategy pages as input, a large language model is used to analyze the types of personal information collected by the target website, the intention of collecting personal information, and the types of third-party services that receive personal information;

[0039] (7) Integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

[0040] In a third aspect, the present invention proposes a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned website data collection strategy analysis method is implemented.

[0041] In a fourth aspect, the present invention provides a computer electronic device comprising a memory and a processor;

[0042] The memory is used to store computer programs;

[0043] The processor is configured to implement the above-mentioned website data collection strategy analysis method when executing the computer program.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) The present invention does not rely on fixed analysis rules and implements a fully automated process from inputting the target website domain name to outputting the analysis report, including automatic crawling of page content, filtering of high-confidence policy pages, and an analysis module based on a large language model. Compared with the unsupervised rule-based methods or fine-tuned classifiers in the prior art, the present invention does not rely on manually predefined rules or fixed analysis frameworks, significantly reducing manual intervention and improving processing efficiency.

[0046] (2) This invention fully leverages the natural language processing advantages of the large language model to achieve intelligent parsing of complex documents. In the analysis module based on the large language model, the large language model is instructed to play the role of an expert, performing analysis of the coverage of website data collection terms and personal data collection practices, thus solving the technical problems of lengthy and complex data collection policy documents. Considering the contextual limitations of the large language model, page segmentation technology is used to ensure rapid processing of large-scale documents, improving overall analysis efficiency while maintaining the consistency of the end-to-end process. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The overall design diagram of the website data collection strategy analysis system based on the large language model;

[0048] Figure 2 Flowchart of the website data collection strategy analysis method based on the large language model. DETAILED DESCRIPTION

[0049] The present invention will be further described and illustrated below in conjunction with specific embodiments. The embodiments are merely illustrative of the present disclosure and do not limit its scope. The technical features of the various embodiments of the present invention may be combined accordingly, provided that there is no conflict between them.

[0050] The accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Some of the blocks shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0051] The present invention provides a website data collection strategy analysis system based on a large language model, which predefines website data collection terms to be analyzed and ultimately generates an analysis result of whether a target website meets the requirements of each predefined term.

[0052] like Figure 1 As shown, the website data collection strategy analysis system mainly includes an input module, a data collection strategy crawling module, a data collection strategy preprocessing module, a website data collection clause coverage analysis module based on a large language model, a website personal data collection practice analysis module based on a large language model, and an output module.

[0053] The functional configuration of each module is introduced as follows:

[0054] (1) Input module

[0055] The input module is used to read the target website domain name and predefined data collection terms of the website to be analyzed.

[0056] The predefined data collection policies that websites must declare include: the types of personal information collected, the types of personal information disclosed, the types of sources of personal information, the purpose of collecting personal information, the purpose of disclosing personal information, the types of third-party services that receive personal information, and the retention period for personal information. Generally, the types of sources of personal information include user-provided information, advertising networks, and data intermediaries; the types of personal information include financial data (credit card data), health information, biometric data (fingerprints), contact information, geographic location, social media data, IP address, computer model, and user online activities; the purposes for collecting or disclosing personal information include providing services (requiring saving your account password for subsequent login verification), advertising, analytical research (to better understand user preferences), customized services, and legal requirements; the types of third-party services that receive personal information include specifically named third-party services (such as Google Analytics), categorized third-party services (such as advertising networks), and public / other users (for example, Weibo users can see your information).

[0057] It should be noted that the types of personal information collected and disclosed do overlap but are not exactly the same. For example, for target website A, the user's biometric data (fingerprint) will be obtained as the login password. This information belongs to the type of personal information collected, but will not be disclosed, that is, it does not belong to the type of personal information disclosed.

[0058] All personal information processing activities involved in the implementation of this invention (including but not limited to financial data, biometric data and other sensitive data) strictly comply with China's current laws, regulations and industry standards.

[0059] The above data collection terms can also be designed according to the needs of technical personnel in this field.

[0060] (2) Data collection strategy crawling module

[0061] The data collection strategy crawling module is used to identify and crawl page contents related to the data collection strategy from the target website as candidate pages.

[0062] It should be noted that a website is a collection of interrelated website pages, media files, and other digital resources (such as images, style sheets, and script files). These files are connected to each other through hyperlinks and share a common root domain name. The present invention targets websites where at least one page contains content related to data collection strategies.

[0063] Specifically, the implementation process of the data collection strategy crawling module includes the following sub-steps:

[0064] (2.1) Extract keywords related to data collection based on the existing multilingual vocabulary, and use keyword matching technology to identify hyperlinks on the target website's homepage that potentially point to the data collection strategy page. The pages pointed to by the hyperlinks are selected as candidate data collection strategy pages.

[0065] (2.2) For the target website homepage and its candidate data collection strategy pages, a web data collection tool such as Selenium is used to load the page, and after it is fully loaded, the page content presented in HTML format is downloaded.

[0066] (2.3) Remove boilerplate content from the page, such as script code, style sheets, and tags (header, footer, nav, etc.). The remaining content is used as the core text content. This text content is then converted into a Markdown format document that is easily understood by the large language model using conversion tools such as html-to-markdown. Subsequent processing is performed on a page-by-page basis.

[0067] (3) Data collection strategy preprocessing module

[0068] The data collection strategy preprocessing module is used to filter candidate pages and extract high-confidence data collection strategy pages.

[0069] Specifically, filtering the candidate data collection strategy pages means: first, for the target website with at least one candidate data collection strategy page, a preliminary screening is performed based on the pre-trained data collection strategy content machine learning classifier; in this embodiment, the Markdown format document corresponding to the candidate data collection strategy page is used as input, and the feature vector is extracted using algorithms such as TextRank and PositionRank, and then the random forest model is used to determine the probability that it belongs to the data collection strategy page. The threshold is set to 0.5, and pages with a probability higher than 0.5 are retained, and the remaining pages are eliminated to obtain a set of pages after preliminary screening.

[0070] Secondly, considering that some terms of service or disclaimer pages may contain content related to data collection policies, which may lead to misjudgment by the classifier, candidate pages whose website page URLs contain keywords such as "data collection" are further selected; here, if the page set after the preliminary screening contains candidate pages with keywords such as "data collection", only these candidate pages are retained for the next operation. If the page set after the preliminary screening does not contain candidate pages with keywords such as "data collection", this process is skipped and the page set after the preliminary screening is directly used for the next operation.

[0071] Finally, considering that the target website may contain third-party service data collection policy descriptions that are unrelated to its own data collection policy or follow the data collection policy of its parent company, the page that meets any of the following criteria is regarded as the final high-confidence data collection policy page: (1) the candidate page whose second-level domain name matches the domain name of the target website; (2) the candidate page whose second-level domain name matches the redirected domain name of the target website.

[0072] The final high-confidence data collection strategy pages are segmented: Taking into account the context window limitations of the large language model, for the obtained high-confidence data collection strategy pages, the Markdown format documents of each page are segmented into fragments that meet the length limitations of the large language model context window, and the operations of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model are executed respectively.

[0073] This module uses a multi-stage filtering operation, first using a pre-trained machine learning classifier for preliminary screening, and then excluding irrelevant pages through URL keyword and second-level domain name matching, ensuring the high reliability of the analysis results, effectively reducing misjudgments, and improving analysis accuracy.

[0074] (4) Website data collection clause coverage analysis module based on large language model

[0075] The website data collection terms coverage analysis module based on the large language model takes the context fragments segmented from the high-confidence data collection policy page as input and uses the large language model to analyze the data collection terms that the target website complies with and does not comply with.

[0076] Specifically, website data collection policy coverage analysis is dedicated to evaluating whether a given data collection policy fragment covers the content required by a certain data collection policy.

[0077] The specific process of analyzing the coverage of website data collection terms based on the large language model includes: first, instructing the large language model to play the role of an expert in in-depth online data collection strategy analysis; second, providing the large language model with a task description, where the task description format is "Your task is to carefully review the provided data collection policy fragment and determine whether it contains the information required by a certain data collection clause", where "a certain data collection clause" needs to be replaced with a preset data collection clause item, such as "explaining the intention to collect consumer personal information"; third, specifying the basic format of the large language model's answer, which should be presented in JSON format and include two aspects: "whether the fragment contains the corresponding data collection clause" and "the description of the fragment corresponding to the corresponding data collection clause"; finally, providing the large language model with a fragment of the segmented target website data collection policy content.

[0078] In this module, the large language model will output the evaluation result of whether a given data collection policy fragment covers the content required by a certain data collection clause, integrate all fragments of all high-confidence data collection policy pages, and take the union of the coverage content of the evaluation results as the evaluation result of the website data collection clause coverage of the target website.

[0079] (5) Website personal data collection practice analysis module based on large language model

[0080] The website personal data collection practice analysis module based on the large language model takes context fragments segmented from high-confidence data collection strategy pages as input and uses the large language model to analyze the types of personal information collected by the target website, the intention of collecting personal information, and the types of third-party services that receive personal information.

[0081] Specifically, the website personal data collection practice analysis is dedicated to evaluating the practices of a given data collection strategy fragment on personal data information, including three categories: the types of personal information collected by the website, the website's intention to collect personal information, and the types of third-party services that receive personal information.

[0082] For each category, the specific process of analyzing website personal data collection practices based on a large language model includes: first, instructing the large language model to play the role of an expert in in-depth online data collection strategy analysis; second, providing the large language model with a task description in the format of "Your task is to carefully review the provided data collection policy fragment and classify a certain personal data collection practice involved in the fragment", where "a certain personal data collection practice" needs to be replaced with a preset personal data collection practice category, such as "the type of personal information collected by the website"; third, providing the large language model with all subordinate classification information under the current preset personal data collection practice category, such as the personal information category that includes categories such as user device identifiers, IP addresses, and session stubs; then, specifying the basic format of the large language model's answer, which should be presented in JSON format and output two aspects for each subordinate classification: "Whether the fragment contains the personal data collection practice of the corresponding subordinate classification" and "The description of the fragment corresponding to the corresponding subordinate classification"; finally, providing the large language model with a fragment of the segmented target website data collection policy content.

[0083] In this module, the large language model will output the evaluation results of each subcategory of the personal information type, the website's intention to collect personal information, and the third-party service type that receives personal information for a given data collection strategy fragment. It will integrate all fragments of all high-confidence data collection strategy pages, take the union of each subcategory of the evaluation results, and use it as the analysis and evaluation results of the target website's personal data collection practices.

[0084] (6) Output module

[0085] The output module is used to integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

[0086] Embodiments of the system of the present invention can be applied to any device with data processing capabilities, such as a computer or other device. System embodiments can be implemented through software, hardware, or a combination of software and hardware. For example, a software implementation, as a logical device, is implemented by a processor of any device with data processing capabilities, reading corresponding computer program instructions from non-volatile memory into internal memory and executing them.

[0087] Based on the same inventive concept, this embodiment also provides a website data collection strategy analysis method based on the above website data collection strategy analysis system, such as Figure 2 As shown, the specific steps include:

[0088] (1) Define the terms of data collection for the website to be analyzed;

[0089] (2) Enter the target website domain name;

[0090] (3) Identify and crawl the page content related to the data collection strategy from the target website as candidate pages;

[0091] (4) Used to filter candidate pages and extract high-confidence data collection strategy pages;

[0092] (5) Using contextual snippets from high-confidence data collection policy pages as input, a large language model is used to analyze the data collection terms that the target website complies with and does not comply with;

[0093] (6) Using contextual fragments extracted from high-confidence data collection strategy pages as input, a large language model is used to analyze the types of personal information collected by the target website, the intention of collecting personal information, and the types of third-party services that receive personal information;

[0094] (7) Integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

[0095] It should also be noted that the website data collection strategy analysis method in the above embodiment can essentially be executed by a computer program. Therefore, similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer electronic device corresponding to the method provided in the above embodiment, which includes a memory and a processor;

[0096] The memory is used to store computer programs;

[0097] The processor is configured to implement the website data collection strategy analysis method in the above embodiment when executing the computer program.

[0098] Furthermore, the logic instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention.

[0099] Therefore, another preferred embodiment of the present invention further provides a computer-readable storage medium corresponding to the method provided in the above embodiment, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the website data collection strategy analysis method in the above embodiment can be implemented.

[0100] It is understood that the storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Furthermore, the storage medium may be any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a magnetic disk, or an optical disk.

[0101] It is understandable that the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0102] It should also be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the various embodiments provided in this application, the division of steps or modules in the system and method is only a logical function division. In actual implementation, there may be other division methods, for example, multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0103] The embodiment described above is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Persons skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent substitution or equivalent transformation falls within the scope of protection of the present invention.

Claims

1. A website data collection strategy analysis system based on a large language model, characterized in that: include: An input module, used to read the target website domain name and predefined data collection terms of the website to be analyzed; The data collection strategy crawling module is used to identify and crawl page contents related to the data collection strategy from the target website as candidate pages; Data collection strategy preprocessing module, used to filter candidate pages and extract high-confidence data collection strategy pages; The website data collection policy coverage analysis module, based on a large language model, takes contextual snippets from high-confidence data collection policy pages as input and uses the large language model to analyze the data collection policies that the target website complies with and does not comply with. A website personal data collection practice analysis module based on a large language model uses contextual snippets extracted from high-confidence data collection strategy pages as input and employs a large language model to analyze the types of personal information collected by the target website, the intention behind collecting personal information, and the types of third-party services that receive personal information. The output module is used to integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

2. The website data collection strategy analysis system based on a large language model according to claim 1 is characterized in that: The predefined data collection terms of the websites to be analyzed include the types of personal information collected, the types of personal information disclosed, the types of sources of personal information, the purpose of collecting personal information, the purpose of disclosing personal information, the types of third-party services that receive personal information, and the retention period of personal information.

3. The website data collection strategy analysis system based on a large language model according to claim 1 is characterized in that: The data collection strategy crawling module is specifically configured as follows: Extract keywords related to data collection terms from a pre-set multilingual vocabulary, and use keyword matching technology to identify hyperlinks on the target website's homepage that potentially point to the data collection policy page. The pages pointed to by the hyperlinks are considered candidate data collection policy pages. Load the candidate data collection strategy page and download the HTML content. Remove the boilerplate content and convert it into a Markdown document. Each candidate data collection strategy page corresponds to a document.

4. The website data collection strategy analysis system based on a large language model according to claim 3 is characterized in that: The data collection strategy preprocessing module is specifically configured as follows: Apply pre-trained machine learning classifiers to initially screen candidate pages; After the initial screening of the results through URL keyword filtering, the candidate pages that meet the URL keyword are retained; if the results after the initial screening do not meet the requirements, the URL keyword filtering process is skipped; Match the target website domain name or redirected domain name based on the second-level domain name, exclude irrelevant pages, and use the remaining results as high-confidence data collection strategy pages; Split the document corresponding to the high-confidence data collection strategy page into segments that meet the constraints of the large language model context window.

5. The website data collection strategy analysis system based on a large language model according to claim 4 is characterized in that: The classification process of the pre-trained machine learning classifier is: Taking the document corresponding to the high-confidence data collection strategy page as input, the feature vector is extracted based on the TextRank or PositionRank algorithm, and the feature vector is input into the random forest model to calculate the probability that the document content meets the data collection strategy page. The pages corresponding to the document content with a probability greater than the threshold are used as candidate pages after preliminary screening.

6. The website data collection strategy analysis system based on a large language model according to claim 1 is characterized in that: The website data collection clause coverage analysis module based on the large language model is specifically configured as follows: Instruct the large language model to play the role of an expert and provide a task description. The task description is to determine whether the input segment contains predefined information about the data collection terms of the website to be analyzed, and output the result in JSON format. Integrate all the fragments of all high-confidence data collection strategy pages, take the union of the terms output by the large language model, and use it as the evaluation result of the website data collection terms coverage of the target website.

7. The website data collection strategy analysis system based on a large language model according to claim 1 is characterized in that: The website personal data collection practice analysis module based on the large language model is specifically configured as follows: Instruct the large language model to act as an expert and provide a task description. The task description is to classify the subcategories of personal data collection practices in the input segment and output a result in JSON format. The personal data collection practices include three categories: the type of personal information collected, the intention of collecting personal information, and the type of third-party services that receive personal information. Each category of personal data collection practices contains several subcategories. Integrate all snippets of all high-confidence data collection strategy pages, take the union of each subcategory output by the large language model, and use it as the analysis and evaluation results of the target website's personal data collection practices.

8. A website data collection strategy analysis method based on the website data collection strategy analysis system according to claim 1, characterized in that: The following steps are involved: (1) Define the terms of data collection for the website to be analyzed; (2) Enter the target website domain name; (3) Identify and crawl the page content related to the data collection strategy from the target website as candidate pages; (4) Used to filter candidate pages and extract high-confidence data collection strategy pages; (5) Using contextual snippets from high-confidence data collection policy pages as input, a large language model is used to analyze the data collection terms that the target website complies with and does not comply with; (6) Using contextual fragments extracted from high-confidence data collection strategy pages as input, a large language model is used to analyze the types of personal information collected by the target website, the intention of collecting personal information, and the types of third-party services that receive personal information; (7) Integrate the analysis results of the website data collection terms coverage analysis module based on the large language model and the website personal data collection practice analysis module based on the large language model, and comprehensively output the data collection strategy analysis report of the target website.

9. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the website data collection strategy analysis method according to claim 8 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the website data collection strategy analysis method according to claim 8 when executing the computer program.

Images