Method of identifying user data access and data deletion rights in an english privacy policy

Abstract

The application discloses a method for identifying user data access and data deletion rights in an English privacy policy. The method mainly utilizes NLP related technologies to gradually extract information content with fine granularity. First, paragraphs belonging to user data access, data editing and data deletion categories are obtained from the privacy policy text. Then, sentences of user data access and user data deletion are obtained from the paragraphs. Next, data access right information, data copy request right information, data deletion right information and account deletion right information are respectively identified. According to the right information, the corresponding right execution mode is identified. Finally, the right execution mode is implemented to obtain the right information of user data access and user data deletion in the privacy policy. The method combines text classification and regular expression matching to identify the right information and the corresponding right execution mode in the sentence, so that the user can understand the data management function provided by the APP through the privacy policy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network security and privacy policy identification, and particularly relates to a method for identifying user data access and data deletion rights in an English privacy policy. BACKGROUND

[0002] A privacy policy is a document provided by a mobile application APP to describe privacy-related terms, which generally includes descriptions of user information collection, third-party information sharing, user data right management, etc. The APP generally collects some information data in the process of use, including user personal information such as name and contact information, and user data generated during use of the APP. The APP uses this data to analyze users and provide personalized services. Users can learn from the privacy policy which information the APP will collect, how the information will be collected, and the third-party information sharing list and other related privacy information. Users can also learn from the privacy policy which data management functions the APP provides to users. Therefore, the privacy policy is a powerful tool for understanding the privacy content of the APP.

[0003] User data management generally refers to the ability of a user to access, edit, and delete their own data while using the APP. Among them, the data access function refers to the ability of a user to understand which information the APP has collected about them while using the APP, as well as the data generated during use of the APP. The APP may provide general data access functions for users, who need to view the corresponding information according to the categories. The APP may also provide data copy access functions, allowing users to request access to all data copies to obtain all current information data, making it easy for users to obtain all data. The data editing function refers to the ability of a user to edit and update incorrect or outdated personal information data while using the APP. The data deletion function refers to the ability of a user to delete information data that they consider too private, or to delete all data after a certain period of use. Or, if a user no longer wants to use the APP, they may want to delete the account and clear all user data. These user data management functions may be performed in various ways, such as directly contacting the enterprise customer service or directly executing in the application.

[0004] The related research on APP privacy policy information extraction at the present stage is mainly about privacy practice, that is, the extraction of information content related to user information collection, such as the extraction of information collection categories, purposes and the like. There are few researches on the identification of user data rights in APP privacy policy, and most of them use text classification to identify the right type declared in the APP privacy policy. Especially for English APP privacy policy, it is difficult to accurately extract relevant information by relying on text classification. Therefore, a method for identifying user data access and data deletion rights in English privacy policy is urgently needed. SUMMARY

[0005] The summary part of the present application is used to introduce the concepts in a brief form, which will be described in detail in the specific embodiment part. The summary part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] In view of the problems and deficiencies in the prior art, the present application aims to provide a method for identifying user data access and data deletion rights in English privacy policy. In the process of information extraction, a gradual fine-grained method is adopted. First, relevant paragraphs are extracted, then relevant sentences are extracted from the extracted paragraphs, the right information contained in the sentences is judged by regular expression matching method, then the execution mode of the right information contained in the extracted relevant sentences is identified, and finally the right information of user data access and user data deletion in the privacy policy is obtained by executing the right execution mode. According to the related technology of NLP, the right information in the sentence and the corresponding right execution mode are identified by combining the method of text classification and regular expression matching, so that the relevant right data information can be extracted from the privacy policy with high accuracy, which is used to solve the problems in the background art.

[0007] To achieve the above object, the present application provides the following technical scheme:

[0008] The present application discloses a method for identifying user data access and data deletion rights in English privacy policy, mainly comprising the following steps:

[0009] Step 1, obtaining the privacy policy text in the mobile application APP, and cutting the privacy policy text to obtain paragraphs belonging to the categories of user data access, user data editing and user data deletion;

[0010] Step 2, cutting the paragraphs of the user data access and user data deletion categories to obtain sentences of user data access and user data deletion;

[0011] Step 3, identifying data access right information and data copy request right information from the sentences accessed by the user data;

[0012] Step 4, identifying data deletion right information and account deletion right information from the sentences deleted by the user data;

[0013] Step 5, identifying the corresponding right execution mode according to the data access right information, data copy request right information, data deletion right information and account deletion right information;

[0014] Step 6, executing the right execution mode to obtain the right information of user data access and user data deletion in the privacy policy.

[0015] Further, in step 1, the privacy policy text in the mobile application APP is obtained, and the specific method is:

[0016] For a plurality of mobile applications APP, the Sprapy framework of python is used to randomly crawl the script content, and the data information of each mobile application APP is obtained, including APP category and privacy policy text.

[0017] Further, in step 1, the privacy policy text is decomposed and cut to obtain the paragraphs belonging to the categories of user data access, user data editing and user data deletion, and the specific steps include:

[0018] Step 1.1, using the existing marked data information about user data access and user data deletion as training sample data;

[0019] Step 1.2, training the TextCNN text multi-classification model according to the training sample data;

[0020] Step 1.3, using the TextCNN text multi-classification model to obtain the categories of all paragraphs for the privacy policy text;

[0021] Step 1.4, retaining the paragraphs belonging to the categories of user data access, user data editing and user data deletion in the paragraphs, and filtering out the categories of the remaining paragraphs.

[0022] Further, in step 2, the paragraphs of user data access and user data deletion are decomposed and cut to obtain the sentences of user data access and user data deletion, and the specific steps include:

[0023] Step 2.1, using artificial extraction of sentences related to user data access and user data deletion from the privacy policy text, and randomly combining irrelevant sentences as training data set;

[0024] Step 2.2, two Bert-based binary text classification models are trained according to the training data set, which are user access sentence extraction model and user deletion sentence extraction model respectively;

[0025] Step 2.3, for each sentence in the paragraphs belonging to the user data access, user data editing and user data deletion categories, the user access sentence extraction model and the user deletion sentence extraction model are used to obtain the related sentences.

[0026] Further, in step 3, data access right information and data copy request right information are identified from the user data access sentences, which are specifically represented as:

[0027] The data access right information is identified from the user data access related sentences whether there is a description of user accessing own data, and the data copy request right information is identified from the user data access related sentences whether there is a description of user requesting own data copy; The identification of the data access right information and the data copy request right information is realized by the method of regular expression matching.

[0028] Further, in step 4, data deletion right information and account deletion right information are identified from the user data deletion sentences, which are specifically represented as:

[0029] The data deletion right information is identified from the user data deletion related sentences whether there is a description of user deleting own data, and the account deletion right information is identified from the user data deletion related sentences whether there is a description of user deleting own account; The data deletion right information is realized by the method of regular expression matching, and the account deletion right information is realized by the method of regular expression matching combined with prior adding part-of-speech tagging and dependency relationship as constraint conditions.

[0030] Further, in step 5, the right execution mode includes contact enterprise customer service setting, APP setting, web page setting and account setting, and the specific execution steps are:

[0031] Step 5.1, judge whether the verbs in the data access right information, data copy request right information, data deletion right information and account deletion right information are identified as contact enterprise customer service setting;

[0032] Step 5.2, if yes, the contact enterprise customer service setting is considered as its right execution mode, if not, continue to find all prepositional phrases in the right information adverbial;

[0033] Step 5.3, by the method of regular expression keyword matching, the corresponding keywords of the right information are searched to determine the corresponding right execution mode;

[0034] Step 5.4, the obtained right execution mode is bound to the right information.

[0035] Compared with the prior art, the present application has the beneficial effects that: the present application provides a method for identifying user data access and data deletion rights in an English privacy policy, which mainly uses NLP related technologies to extract relevant information content by gradually fine-grained method. First, the paragraphs belonging to the categories of user data access, data editing and data deletion are obtained, and then the sentences of user data access and user data deletion are obtained based on the paragraphs, and the data access right information, data copy request right information, data deletion right information and account deletion right information are identified respectively, and then the right execution mode corresponding to the right information is identified, and finally the right execution mode is executed to obtain the right information of user data access and user data deletion in the privacy policy. According to the grammatical characteristics of the English privacy policy text, the right information and the corresponding right execution mode in the sentence are identified by combining the text classification and regular expression matching method, which can extract relevant information from the English privacy policy text with high accuracy, and it is convenient to understand the data management function provided by the APP to the user through the privacy policy. BRIEF DESCRIPTION OF DRAWINGS

[0036] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrations, together with their description, serve to explain the application.

[0037] In the drawings:

[0038] Figure 1 : the main flow structure diagram of the method for identifying user data access and data deletion rights in the implemented English privacy policy of the present application;

[0039] Figure 2 : the main step flow diagram of the method for identifying user data access and data deletion rights in the implemented English privacy policy of the present application;

[0040] Figure 3 : the flow structure diagram of the right execution mode identification in the method for identifying user data access and data deletion rights in the implemented English privacy policy of the present application;

[0041] Figure 4 : the dependency relationship structure diagram of the added account deletion information sentence in the method for identifying user data access and data deletion rights in the implemented English privacy policy of the present application. DETAILED DESCRIPTION

[0042] Embodiments of the present disclosure will be described below in greater detail with reference to the accompanying drawings. While certain embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thoroughly and completely understood. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0043] In addition, it should be further noted that, for ease of description, only the parts related to the present application are shown in the drawings. The embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0044] The present disclosure discloses a method for identifying user data access and data deletion rights in an English privacy policy, which will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0045] The method of the present disclosure is mainly realized by using NLP related technology. NLP is a technology for computer to automatically process natural language, including text classification, semantic recognition, etc. The first step of NLP processing natural text is generally to represent the text in a way that can be understood by the computer, i.e. to vectorize the text. There are many ways to represent text vectors, such as bag-of-words model, topic vector, word vector, etc. Referring to Figures 1 to 2 The main steps include:

[0046] Step 1, obtaining the privacy policy text in the mobile application APP, and cutting the privacy policy text to obtain the paragraphs belonging to the user data access, user data editing and user data deletion categories;

[0047] Step 2, cutting the paragraphs of the user data access and user data deletion categories to obtain the sentences of the user data access and user data deletion;

[0048] Step 3, identifying the data access right information and the data copy request right information from the sentences of the user data access;

[0049] Step 4, identifying the data deletion right information and the account deletion right information from the sentences of the user data deletion;

[0050] Step 5, identifying the corresponding right execution mode according to the data access right information, the data copy request right information, the data deletion right information and the account deletion right information;

[0051] Step 6, executing the right execution mode to obtain the right information of the user data access and user data deletion in the privacy policy.

[0052] Specifically, the privacy policy is presented to the user, which is a text describing privacy-related terms. Most of the privacy policy text is long and contains many paragraphs, and describes various types of privacy terms, including user data collection, third-party information sharing, user data use, etc. Among them, the content used to describe user data access and data deletion functions belongs to the user data right category, which only occupies a small part of the privacy policy text, and the vast majority of the content of the privacy policy text is not of interest to us, so it needs to be filtered out. Therefore, the present application adopts a gradual fine-grained method in the information extraction process. First, the privacy policy text is obtained, and the paragraphs belonging to the user data access, data editing and data deletion categories are obtained by cutting the privacy policy text. On the basis of the paragraph, the user data access and user data deletion sentences are obtained, and the data access right information, data copy request right information, data deletion right information and account deletion right information are respectively identified. Then the right execution mode corresponding to the right information is identified, and finally the right execution mode is executed to obtain the right information of user data access and user data deletion in the privacy policy.

[0053] The privacy policy text in the mobile application APP is obtained, specifically: a plurality of mobile applications APP are selected, and 941 script contents are randomly crawled through the Sprapy framework of python to correspondingly obtain 941 data information of the mobile applications APP. The data information includes the category of the APP and the privacy policy text. The Scrapy crawler framework is an encapsulated framework, which includes a downloader, a parser, a log and an exception handling, and is processed based on multi-threading in the twisted mode. Sprapy adopts a more readable XPath instead of a regular expression, can crawl on different URLs at the same time, and is convenient for independent debugging.

[0054] In step 1, the privacy policy text is decomposed and cut to obtain paragraphs belonging to the user data access, user data editing and user data deletion categories, and the specific steps include:

[0055] Step 1.1, using the existing marked data information about user data access and user data deletion as training sample data;

[0056] Step 1.2, training the TextCNN text multi-classification model according to the training sample data;

[0057] Step 1.3, using the TextCNN text multi-classification model to obtain the categories of all paragraphs for the privacy policy text;

[0058] Step 1.4, retaining the paragraphs belonging to the user data access, user data editing and user data deletion categories in the paragraphs, and filtering out the categories of the remaining paragraphs.

[0059] Specifically, the paragraph extraction uses a TextCNN text multi-classification model, which is a convolutional neural network for text classification. The training sample data of this model comes from existing privacy policy research. The text content in the privacy policy is divided into 12 categories according to the granularity of the paragraph, as shown in Table 1 below:

[0060]

[0061] Table 1

[0062] As can be seen from the table, the paragraphs in the user data access, user data editing and user data deletion categories are the paragraphs that need to be extracted, corresponding to the User Access, Edit and Deletion paragraph labels in the table. By using existing labeled data about user data access and user data deletion as training sample data, a TextCNN text multi-classification model is trained. For each privacy policy text obtained, some processing is first performed, and then this model is used to classify each paragraph in the privacy text. Only the paragraphs belonging to the user data access, user data editing and user data deletion categories are retained, and the remaining paragraphs that we do not need are filtered out.

[0063] Further, the convolutional neural network CNN is a kind of feedforward neural network model, which is widely used in image processing. TextCNN is a CNN model applied to text classification tasks, which can also achieve good text classification results. In each paragraph of the privacy policy text, it is determined whether it belongs to the user data access, user data editing and user data deletion category paragraph, which is a binary classification task. However, we use a TextCNN text multi-classification model to identify it. This is because if a binary classification model is trained, the problem of unbalanced positive and negative samples is likely to occur, resulting in poor recognition effect of the model. Therefore, using a multi-classification model can well solve this problem and achieve good text classification results.

[0064] In step 2, the user data access and user data deletion category paragraphs are decomposed and cut to obtain user data access and user data deletion sentences, and the specific steps include:

[0065] Step 2.1, use artificial extraction of sentences related to user data access and user data deletion from the privacy policy text, and randomly combine irrelevant sentences as training data set;

[0066] Step 2.2, two two-part text classification models based on Bert are trained according to the training data set, which are user access sentence extraction model and user deletion sentence extraction model;

[0067] Step 2.3, for each sentence belonging to the paragraph in the category of user data access, user data editing and user data deletion, use the user access sentence extraction model and the user deletion sentence extraction model to obtain the relevant sentences.

[0068] Specifically, after extracting the paragraphs belonging to the categories of user data access, user data editing and user data deletion from the privacy policy text, it is necessary to continue to extract relevant sentences from these paragraphs. Here we divide the sentences into two categories, one related to user data access and the other related to user data deletion, and train two binary text classification models based on Bert to identify the two categories of sentences. Since there is no existing data available to train the two classification models, it is necessary to manually find some sentences related to user data access and user data deletion from the privacy policy text, and then randomly combine some irrelevant sentences as training data sets, and then train two binary text classification models based on Bert, namely the user access sentence extraction model and the user deletion sentence extraction model. The former is used to identify whether there is a sentence related to user data access, and the latter is used to identify whether there is a sentence related to user data deletion, obtaining user data access sentences and user data deletion sentences.

[0069] Further, using the user access sentence extraction model and the user deletion sentence extraction model to obtain relevant sentences, we can also use a simple neural network classification model based on Bert text embedding to achieve this. This simple neural network classification model can be implemented through a fully connected layer and an activation function. Since this neural network classification model is very simple, the advantages of Bert can achieve good classification results. Bert is also a new type of word vector representation model, which has good results for NLP related tasks. Text classification is to divide text into different categories according to its type, which can be achieved in many ways, including traditional machine learning and deep neural networks. Bert is a pre-trained language representation model that emphasizes no longer using traditional one-way language models or shallowly splicing two one-way language models for pre-training, but using a new method to generate deep bidirectional language representation.

[0070] In step 3, data access right information and data copy request right information are identified from the user data access related sentences. The data access right information is identified from the user data access related sentences whether there is a description of the user's access to his own data, and the data copy request right information is identified from the user data access related sentences whether there is a description of the user's request for his own data copy; the identification of data access right information and data copy request right information is realized by the method of regular expression matching.

[0071] Specifically, the data access right information is obtained by manually observing some expressions of data access, and there are generally two common expression modes. One is the active voice: verb + noun, and the other is the passive voice: noun + verb. The verb is an expression of access-related meaning, such as access, review, etc. The noun is an expression of information data, such as data, information, etc. Using these two expression modes and using the regular expression matching method, it can be determined whether the sentence contains a description that the user can access his own data.

[0072] The data copy request right information is obtained by manually observing some expressions of data copy request, and some common expression modes are found as follows: one is a noun phrase, such as: copy of information / data, data portability. The other is a verb-object structure, which also has two forms of active and passive voice, the verb has download and export, and the noun is also an expression of information data. By combining these expression modes and using the regular expression matching method, it can be determined whether the sentence contains a description that the user can request a copy of his own data.

[0073] In step 4, the data deletion right information and the account deletion right information are identified from the user data deletion related sentences. The data deletion right information is identified from the user data deletion related sentences whether there is a description that the user can delete his own data, and the account deletion right information is identified from the user data deletion related sentences whether there is a description that the user can delete his own account. The data deletion right information is realized by the regular expression matching method, and the account deletion right information is realized by adding part-of-speech tags and dependency relations as constraints in advance, and then combining the regular expression matching method.

[0074] Specifically, the data deletion right information is obtained by manually observing some expressions of data deletion, and it is found that the matching mode is similar to the user data access category. There are two expression modes of active and passive voice of verb-object structure, the verb includes delete, erase, etc., and the noun is also an expression of information data. In addition, through observation, there is another common expression mode, i.e. the nominalization form of the verb, such as deletion of, erasure of, etc., and of is followed by a noun expressing information data. By comprehensively using these expression modes and using the regular expression matching method, it can be determined whether the sentence contains a description that the user can delete his own data.

[0075] The account deletion right information is found to have a single expression mode by manually observing some expressions describing account deletion, which is generally a verb-object structure, with the verb including delete, terminate, etc., and the object being the noun account. However, if this expression mode is directly used in combination with regular expression matching, there will be more false matches. For example, there can be expressions such as delete…data of account in the data deletion description, which means deleting data. However, this expression will be identified as deleting an account when matched with the regular expression matching mode, which is a false match. Therefore, the part-of-speech tagging and dependency relationship information need to be added as a constraint condition when using regular expression matching to identify the deletion of account information, and it is also necessary to ensure that the noun account is the object of the preceding verb. With this restriction, false matches can be excluded.

[0076] The part-of-speech information and dependency relationship are added as follows: first, determine the relevant verbs, which are determined based on the part-of-speech tagging of the verbs as VERB. Then, determine the object based on the dependency relationship dobj. Due to the existence of parallel verb structures, account may not be the direct object of the corresponding verb, but the object of the last verb in the parallel verb structure. Therefore, it is necessary to traverse the parallel verbs to determine the subject-object relationship. We use spacy to perform part-of-speech tagging and dependency analysis on sentences. Spacy is a popular NLP processing tool library. Part-of-speech tagging refers to the tagging of the part of speech of words in a sentence, such as verbs, nouns, etc. Dependency relationship refers to the relationship between words, such as the relationship between verbs and objects, subjects and verbs, etc. The result of dependency analysis is a dependency relationship tree, with the verb as the root structure. An example of dependency relationship is shown in Figure 4 Based on these expression modes and methods, it is determined whether the sentence contains a description of the user's ability to delete his / her own account.

[0077] The regular expression matching method used for the data access right information, the data copy request right information, the data deletion right information, and the account deletion right information has a matching mode for each type of right information as shown in Table 2:

[0078]

[0079] Table 2

[0080] Referring to Figure 3 As shown in FIG. 5, the right execution mode in step 5 includes contacting enterprise customer service settings, APP settings, web page settings, and account settings, and the specific steps of the right identification mode are as follows:

[0081] Step 5.1, determine whether the verb in the data access right information, data copy request right information, data deletion right information and account deletion right information is identified as contacting enterprise customer service setting;

[0082] Step 5.2, if yes, consider contacting enterprise customer service setting as its right execution mode, if not, continue to find all prepositional phrases in the right information adverbial;

[0083] Step 5.3, by regular expression keyword matching method, find the corresponding keywords of the right information to determine its corresponding right execution mode;

[0084] Step 5.4, bind the obtained right execution mode with the right information.

[0085] Specifically, there are two common sentence patterns in the privacy policy, one is that the right execution mode is in the manner adverbial, and the other is that the action in the sentence represents the right execution mode. Contacting enterprise customer service setting is the most common user data right execution mode, in addition to contacting enterprise customer service setting, there are app setting, web setting and account setting. We found that the sentence pattern of the latter three right execution modes in the privacy policy is mainly in the form of manner adverbial, we observed the manner adverbial expression of each right execution mode, and summarized the keyword table of each right execution mode, the keyword table information is shown in Table 3:

[0086]

[0087] Table 3

[0088] According to the above content, we identify the right execution mode information in the sentence according to the following method. For a sentence containing data right description, we first determine whether its verb is contact or email, which are both indicating contacting enterprise customer service setting. If the verb is one of the two verbs, we consider that the sentence contains the description of contacting enterprise customer service as the right execution mode. If the verb does not meet the matching conditions, we need to continue to find all prepositional phrases in the adverbial of the sentence, for each prepositional phrase, we use regular expression keyword matching method to find whether the corresponding keyword of each right execution mode is in it. If it exists, it is considered that the sentence contains the corresponding right execution mode information. Finally, the identified right execution mode in the sentence is bound with the data right information extracted from the sentence.

[0089] After determining how to identify each type of right information and the right execution method, we can identify the data right information contained in the sentence. For the sentence of user data access, we can first match the data access right information, and then match the data copy request right information. For the sentence of user data deletion, we can first match the data deletion right information, and then match the account deletion right information. The right information of the privacy policy text after sentence extraction is extracted, if the sentence contains the description of the data right information, the right execution method identification method is used to identify the right execution method statement contained in the sentence, and the right information of the user data access and the user data deletion of the entire privacy policy statement can be obtained.

[0090] Experimental results

[0091] In order to verify the accuracy of the method, 80 mobile applications APPs were manually selected, and their privacy policies were read. For each privacy policy, four types of right information, including data access right information, data copy request right information, data deletion right information, and account deletion right information, and execution method information were extracted, and the information of the 80 mobile applications APPs was summarized as test data. The method was used to automatically extract information from the privacy policies of the APPs, and the extraction results were compared with the test data.

[0092] The data right type recognition accuracy experimental results are shown in Table 4 as follows:

[0093]

[0094] Table 4

[0095] The right execution method recognition accuracy results are shown in Table 5 as follows:

[0096]

[0097] Table 5

[0098] The above description is only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or equivalent features without departing from the above inventive concept. For example, the above features are replaced with the technical features disclosed in the embodiments of the present disclosure (but not limited to) having similar functions to form technical solutions.

Claims

1. A method of identifying user data access and data deletion rights in an English privacy policy, characterized by, Mainly includes the following steps: Step 1, obtain the privacy policy text in the mobile application APP, the privacy policy text is obtained by decomposition cutting the paragraphs belonging to user data access, user data editing and user data deletion category; Step 2, the paragraph of user data access and user data deletion category is decomposed and cut to obtain the sentence of user data access and user data deletion; Step 3, the data access right information and the data copy request right information are identified from the sentence of user data access; Step 4, the data deletion right information and the account deletion right information are identified from the sentence of user data deletion; Step 5, the right execution mode corresponding to the data access right information, the data copy request right information, the data deletion right information and the account deletion right information is identified; Step 6, the right execution mode is executed to obtain the right information of user data access and user data deletion in the privacy policy; The privacy policy text is decomposed and cut to obtain the paragraph belonging to user data access, user data editing and user data deletion category, and the specific steps include: Step 1.1, using the existing marked data information about user data access and user data deletion as training sample data; Step 1.2, training the TextCNN text multi-classification model according to the training sample data; Step 1.3, using the TextCNN text multi-classification model to obtain the category of all paragraphs for the privacy policy text; Step 1.4, retaining the paragraphs belonging to user data access, user data editing and user data deletion category in the paragraph, and filtering out the category of the remaining paragraphs; The paragraph of user data access and user data deletion category is decomposed and cut to obtain the sentence of user data access and user data deletion, and the specific steps include: Step 2.1, using artificial extraction of sentences related to user data access and user data deletion from the privacy policy text, and randomly combining irrelevant sentences as training data set; Step 2.2, two two-part text classification models based on Bert are trained according to the training data set, which are user access sentence extraction model and user deletion sentence extraction model; Step 2.3, for each sentence in the paragraph belonging to user data access, user data editing and user data deletion category, the user access sentence extraction model and the user deletion sentence extraction model are used to obtain the related sentences.

2. The method of identifying user data access and data deletion rights in an English privacy policy of claim 1, wherein, In step 1, the privacy policy text in the mobile application APP is obtained, and the specific method is: for a plurality of mobile applications APP, the Sprapy framework of python is used to randomly crawl the script content, and the data information of each mobile application APP is obtained, the data information of the mobile application APP includes APP category and privacy policy text.

3. The method of identifying user data access and data deletion rights in an English privacy policy of claim 2, wherein, In step 3, the data access right information and the data copy request right information are identified from the sentence of user data access, which are specifically represented as: The data access right information is identified from the user data access related sentence whether there is a description of the user accessing his own data, and the data copy request right information is identified from the user data access related sentence whether there is a description of the user requesting his own data copy; The identification of the data access right information and the data copy request right information is realized by the method of regular expression matching.

4. The method of identifying user data access and data deletion rights in an English privacy policy of claim 3, wherein, The data deletion right information and the account deletion right information are identified from the sentence deleted by the user data in step 4, which is specifically represented as: The data deletion right information is identified from the user data deletion related sentence whether there is a description of the user deleting his own data, and the account deletion right information is identified from the user data deletion related sentence whether there is a description of the user deleting his own account; The data deletion right information is realized by the method of regular expression matching, and the account deletion right information is realized by the method of regular expression matching combined with the constraint conditions of part-of-speech tagging and dependency relationship in advance.

5. The method of identifying user data access and data deletion rights in an English privacy policy of claim 4, wherein, The right execution mode in step 5 includes contact enterprise customer service setting, APP setting, web page setting and account setting, and the specific steps of the right execution mode are: Step 5.1, judge whether the verb in the data access right information, data copy request right information, data deletion right information and account deletion right information is identified as contact enterprise customer service setting; Step 5.2, if yes, the contact enterprise customer service setting is considered as the right execution mode, and if not, continue to find all preposition phrases in the right information adverbial; Step 5.3, by the method of regular expression keyword matching, the corresponding keywords of the right information are found to determine the corresponding right execution mode; Step 5.4, the right execution mode obtained is bound with the right information.

Images