Method and electronic device for detecting personal information

Abstract

Provided is a method for detecting personal information, performed by an electronic device. The method may comprise the steps of: obtaining first text of a first language; detecting personal information from the first text; translating the first text into a second language to obtain second text; calculating a first score on the basis of a phonetic analysis of the personal information and the second text; calculating a second score on the basis of a semantic analysis of the personal information and the second text; and detecting personal information elements corresponding to the personal information of the first text from the second text on the basis of the first score and the second score.

Description

Personal Information Detection Method and Electronic Device

[0001] The present disclosure relates to an electronic device and method for detecting personal information elements within text using a single cross-language model capable of detecting personal information in several different languages.

[0002] As interest in and the importance of personal information protection increase, regulations to safeguard it are continuously being introduced, and various technologies for this purpose are being developed simultaneously. In particular, the importance of technologies capable of accurately detecting personal information within text is growing. However, since personal information is expressed differently depending on the language, problems arise where detection becomes difficult or new detection methods must be applied every time the language changes. This inefficiency becomes even more severe as the scale of data increases. Therefore, there is a growing need for methods that can efficiently detect the same personal information across various languages, regardless of the specific language.

[0003] According to one aspect of the present disclosure, a method for detecting personal information performed by an electronic device may be provided. The method may include the steps of: obtaining a first text of a first language; detecting personal information in the first text; translating the first text into a second language to obtain a second text; identifying a first score based on a pronunciation analysis of the personal information and the second text; identifying a second score based on a semantic analysis of the personal information and the second text; and detecting personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

[0004] According to one aspect of the present disclosure, an electronic device for detecting personal information may be provided. The electronic device may include a communication interface, at least one processor, and a memory for storing instructions. By executing the instructions by the at least one processor, the electronic device may acquire a first text of a first language, detect personal information in the first text, translate the first text into a second language to acquire a second text, identify a first score based on a pronunciation analysis of the personal information and the second text, identify a second score based on a semantic analysis of the personal information and the second text, and detect personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

[0005] According to one aspect of the present disclosure, a computer-readable recording medium may be provided having a program recorded thereon for executing any one of the methods described above and below, in which an electronic device detects personal information.

[0006] FIG. 1 is a diagram illustrating, by way of example, an operation in which an electronic device according to one embodiment of the present disclosure detects cross-language personal information.

[0007] FIG. 2 is a flowchart illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0008] FIG. 3 is a diagram illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0009] FIG. 4 is a diagram illustrating the operation of a pronunciation analysis module of a cross-language model according to one embodiment of the present disclosure.

[0010] FIG. 5 is a diagram illustrating the operation of a semantic analysis module of a cross-language model according to one embodiment of the present disclosure.

[0011] FIG. 6 is a diagram illustrating the operation of an electronic device according to one embodiment of the present disclosure to detect personal information by combining pronunciation analysis and semantic analysis.

[0012] FIG. 7 is a diagram illustrating an example of an operation in which an electronic device according to one embodiment of the present disclosure detects personal information across languages.

[0013] FIG. 8 is a diagram illustrating an example of an operation in which an electronic device according to one embodiment of the present disclosure detects personal information across languages.

[0014] FIG. 9 is a flowchart illustrating the operation of an electronic device according to one embodiment of the present disclosure applying settings for personal information detection.

[0015] FIG. 10 is a flowchart illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0016] FIG. 11 is a block diagram illustrating the configuration of an electronic device according to one embodiment of the present disclosure.

[0017] FIG. 12 is a block diagram illustrating the configuration of a server according to one embodiment of the present disclosure.

[0018] The terms used in this specification will be briefly explained, and the present disclosure will be described in detail. In the present disclosure, the expression "at least one of a, b, or c" may refer to "a," "b," "c," "a and b," "a and c," "b and c," "all of a, b, and c," or variations thereof.

[0019] The terms used in this disclosure have been selected to be as widely used and general as possible, taking into account their functions within this disclosure; however, these terms may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant explanatory sections. Therefore, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the overall content of this disclosure.

[0020] Singular expressions may include plural expressions unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as generally understood by those skilled in the art as described in this specification. Additionally, terms including ordinal numbers, such as "first" or "second," used in this specification may be used to describe various components, but said components should not be limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another.

[0021] When a part of a specification is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "part" or "module" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware or software, or as a combination of hardware and software.

[0022] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present disclosure in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0023] The present disclosure will be described below with reference to the attached drawings.

[0024] FIG. 1 is a diagram illustrating, by way of example, an operation in which an electronic device according to one embodiment of the present disclosure detects cross-language personal information.

[0025] In one embodiment, the electronic device may detect personal information from text using a cross-language model (100). Personal information includes name, gender, date of birth, address, etc., and may refer to any type of information for identifying an individual.

[0026] Even if the personal information has the same content, the method of expression may differ depending on the language. Therefore, when the language changes, separate language models (101) are required to detect personal information in that language. For example, an English model may be used to detect personal information written in English, a Korean model may be used to detect personal information written in Korean, and a Spanish model may be used to detect personal information written in Spanish.

[0027] The cross-language model (100) of the present disclosure supports a cross-language personal information detection function that can detect personal information regardless of the language of the input text. For example, “Hong Gil-dong” is a Korean name written in Korean, “James” is an English name written in Korean, and “John” is an English name written in English. The cross-language model (100) can detect personal information elements within text written in different languages ​​without changing the model. To this end, the cross-language model (100) can perform tasks to detect the corresponding personal information in another language as is, using additional information that reflects the characteristics of the personal information based on the personal information elements detected in one language.

[0028] In one embodiment, the cross-language model (100) may perform a pronunciation analysis to reflect the characteristic that the pronunciation of personal information is similar even if the language changes. Additionally, the cross-language model (100) may perform a semantic analysis to reflect the characteristic that the meaning of personal information is similar even if the word order changes due to the language change. The cross-language model (100) can detect the same personal information in another language based on personal information identifiable in a specific language by using a combination of pronunciation analysis and semantic analysis. In the present disclosure, a specific language may be referred to as the 'source language' and another language as the 'target language,' and for distinction, the source language and the target language may be referred to as the first language and the second language, respectively.

[0029] In one embodiment, the electronic device may be a device capable of performing a personal information detection task and displaying and providing text and / or personal information detection results. For example, the electronic device may be implemented as an electronic device of various types and forms including a display. The electronic device may include, but is not limited to, devices capable of displaying visual information through a display, such as a smart TV, smartphone, tablet PC, or laptop PC.

[0030] In one embodiment, the electronic device may be a device that performs a personal information detection task and provides text and / or personal information detection results to a user device. For example, the electronic device and the user device of the present disclosure may be implemented in a server-client device structure.

[0031] Specific operations of an electronic device detecting and providing personal information in multiple languages ​​using a cross-language model (100) will be described in more detail through the drawings and descriptions thereof.

[0032] FIG. 2 is a flowchart illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0033] An electronic device according to one embodiment can perform a cross-language personal information detection task, which involves detecting personal information within text written in a specific language and detecting personal information of the same content within text translated from that language into another language.

[0034] In operation S210, the electronic device may obtain the original text of the source language. The original text may be used as input data for a cross-language personal information detection operation performed by the electronic device. The original text of the source language may refer to text written in a specific language (e.g., Korean, English, Chinese, etc.).

[0035] In one embodiment, the electronic device can obtain the original text based on user input. For example, the electronic device can obtain the original text by a method in which the user directly inputs text through a keyboard or voice recognition.

[0036] In one embodiment, the electronic device may acquire text that is input or output during interaction with a user as input text for a personal information detection task. For example, while providing or receiving a chatbot service, the electronic device may acquire text corresponding to a user's question or text corresponding to a chatbot's response as original text.

[0037] In one embodiment, the electronic device may obtain original text from a text file. For example, the electronic device may obtain text files of various types (e.g., txt, csv, JSON, etc.) and read text from the file. The text file may be, for example, a secure document containing personal information, a record of a conversation between a user and a chatbot, but is not limited thereto.

[0038] In operation S220, the electronic device can detect personal information from the original text.

[0039] Personal information may refer to any type of information used to identify an individual. Personal information may include, for example, name, gender, date of birth, address, email address, phone number, social media account, resident registration number, passport number, driver's license number, etc., but is not limited thereto. For convenience of explanation, name and address will be described as primary examples of personal information in this disclosure. However, all content described in this disclosure may be applied equally or similarly to all types of personal information elements.

[0040] In one embodiment, the electronic device can detect personal information in the original text using a rule-based detection method. For example, the electronic device can detect personal information within the original text based on keywords or patterns, such as names included in a name database, phone number formats, email address formats, address formats, etc.

[0041] In one embodiment, an electronic device may detect personal information from original text using a deep learning-based detection method. For example, the electronic device may use an artificial intelligence model based on an artificial intelligence architecture for processing text (e.g., BERT, LSTM, etc.). The electronic device may collect data from various documents containing personal information (e.g., text, email, conversation logs, etc.) and train an artificial intelligence model to detect personal information from text based on the collected data. The detection model for detecting personal information may be an artificial intelligence model trained to detect personal information in the source language, and a cross-language model described in this disclosure may additionally perform the function of the detection model.

[0042] In one embodiment, an electronic device can detect personal information in original text based on user input. The electronic device can receive user input that designates personal information within the text. For example, the electronic device can receive user input that designates the word “Hong Gil-dong” within the text as personal information representing a name.

[0043] In operation S230, the electronic device can obtain the translated text by translating the original text into the target language.

[0044] In one embodiment, the target language may be defined as a default value and refers to a language different from the source language. For example, if the source language is Korean, the target language may be defined as English, Chinese, etc. Or, if the source language is English, the target language may be defined as a language other than English.

[0045] In one embodiment, target language information can be obtained based on user input. For example, if the source language is Korean, the target language can be determined by user input selecting the target language, such as English or Chinese.

[0046] An electronic device can translate original text into a target language. The electronic device can obtain a translated text in which the original text is translated into a target language by using machine translation or deep learning-based translation. In the present disclosure, expressions such as a first text of a first language (e.g., original text of a source language) and a second text of a second language (e.g., translated text of a target language) may be used to distinguish texts written in different languages.

[0047] In one embodiment, the electronic device can translate the original text into an intermediate language. The electronic device can translate the original text into an intermediate language and then translate the text of the intermediate language into a target language. Translation into an intermediate language may be used when direct translation from the original language to the target language is impossible, or when the context of the original language is reflected more naturally in the target language when passing through an intermediate language. For example, the electronic device may include an intermediate language translation process for defined language types based on language identification information.

[0048] In operation S240, the electronic device can calculate a first score based on the pronunciation analysis of personal information and translated text.

[0049] In one embodiment, the electronic device may perform pronunciation analysis to detect corresponding personal information in texts of different languages. Due to the nature of the information, the pronunciation of personal information may be identical or similar even if the language changes. The electronic device may analyze the pronunciation of each text in different languages ​​to detect words or sentences that have similar pronunciations between the texts. For example, 'Personal Information A' may be included in the first text of the first language (source language) and the second text of the second language (target language) in their respective languages. The electronic device may detect words or sentences that have similar pronunciations through the pronunciation analysis of the first text and the second text. The electronic device may calculate a first score indicating the degree of similarity in pronunciation with the personal information within the original text. The specific operation of the electronic device performing pronunciation analysis is described in the description of FIG. 4.

[0050] In operation S250, the electronic device can calculate a second score based on semantic analysis of personal information and translated text.

[0051] In one embodiment, the electronic device may perform semantic analysis to detect corresponding personal information in texts of different languages. When text is translated, a word or sentence that was personal information in the original text may also be personal information in the translated text. The electronic device may analyze the semantics of each text in different languages ​​to detect words or sentences that have similar meanings between the texts. For example, 'Personal Information A' may be included in the first text of the first language (source language) and the second text of the second language (target language) in their respective languages. The electronic device may detect words or sentences that have associations between each text through semantic analysis of the first text and the second text. The electronic device may calculate a second score indicating semantic association with personal information within the original text. The specific operation of the electronic device performing semantic analysis is described in the description of FIG. 5.

[0052] In operation S260, the electronic device can detect personal information elements corresponding to the personal information of the original text in the translated text based on the first score and the second score.

[0053] In one embodiment, the electronic device can identify candidate sentences or words in the translated text that have pronunciations similar to personal information in the original text based on a first score. For example, the electronic device can sort sentences or words in the translated text in order of the first score and identify the top N sentences or words as candidates for personal information elements. Alternatively, for example, the electronic device can identify sentences or words whose first score exceeds a threshold as candidates for personal information elements.

[0054] In one embodiment, the electronic device can identify candidate sentences or words in the translated text that have a meaning similar to personal information within the original text based on a second score. For example, the electronic device can sort sentences or words within the translated text in order of the second score and identify the top M sentences or words as candidates for personal information elements. Alternatively, for example, the electronic device can identify sentences or words whose second score exceeds a threshold as candidates for personal information elements.

[0055] The electronic device can determine a final personal information element from among the candidates for identified personal information elements based on the first score and the second score. The final personal information element may be a sentence or word within the translated text corresponding to the personal information of the original text.

[0056] For example, the electronic device may determine a common element as a personal information element among a first personal information element candidate determined by a first score and a second personal information element candidate determined by a second score. Alternatively, for example, the electronic device may determine the element having the highest overall score as a personal information element based on the sum of the first score and the second score.

[0057] In one embodiment, the electronic device may apply weights based on language characteristics and personal information characteristics. In one embodiment, the electronic device may use a combination of weights based on language characteristics and weights based on personal information characteristics. The electronic device may detect personal information elements corresponding to the personal information of the original text in the translated text based on the weighted sum of the first score and the second score to which the weights are applied. The operation of the electronic device applying weights is further described in the description of FIG. 6.

[0058] In one embodiment, the electronic device may determine the priority between pronunciation analysis and semantic analysis by comparing at least one of a first score or a second score with a threshold. For example, the electronic device may set the priority of pronunciation analysis high if there is a word whose first score is above the threshold as a result of pronunciation analysis, and the first scores of the remaining words are below the threshold. In other words, if it is determined during the analysis process that only one word has high pronunciation similarity and the reliability of the pronunciation analysis result is sufficiently high, pronunciation analysis may be applied preferentially. If the priority of pronunciation analysis is determined to be high, the electronic device applies pronunciation analysis preferentially in subsequent personal information detection operations, and may omit semantic analysis if the reliability of pronunciation analysis is sufficiently high (e.g., above the threshold). In the same way, the priority of semantic analysis may be adjusted to be high. In this case, if the priority of semantic analysis is determined to be high, the electronic device applies semantic analysis preferentially in subsequent personal information detection operations, and may omit pronunciation analysis if the reliability of semantic analysis is sufficiently high.

[0059] In one embodiment, the electronic device may determine the order of pronunciation analysis and semantic analysis. For example, if one analysis method is easier to use than another, the easier method may be applied preferentially. Specifically, if it is difficult to perform pronunciation analysis due to the lack of pronunciation information for a specific language, the electronic device may apply semantic analysis preferentially. Additionally, for example, if one analysis method demonstrates higher performance than another, the analysis method with higher performance may be applied preferentially. Such analysis performance may vary depending on the characteristics of the language or personal information. The priority may be adjusted by weights.

[0060] FIG. 3 is a diagram illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0061] In one embodiment, the cross-language model (300) can perform a personal information detection task for the cross-language. For example, the cross-language model (300) can detect personal information for various target languages ​​based on personal information detected for the source language. To explain with the example where the source language is 'Language A', the cross-language model (300) can receive text written in Language A as input and detect personal information elements. The cross-language model (300) may include a detection module trained to enable personal information detection for the source language, Language A. In this case, the detection module may be based on an artificial intelligence architecture for processing text (e.g., BERT, LSTM, etc.). Alternatively, the cross-language model (300) may obtain information about personal information elements based on user input specifying personal information elements for Language A.

[0062] The cross-language model (300) can detect personal information elements within text written in a language different from the source language based on information about personal information elements of the source language. To explain with an example where the target language is 'language B', the cross-language model (300) can receive text written in the target language, language B, as input and detect personal information elements. Language B can be any one or more languages ​​different from language A. For example, if language A is Korean, language B can be any languages ​​different from Korean, such as English, Chinese, Japanese, etc.

[0063] In one embodiment, the cross-language model (300) may include a pronunciation analysis module (310) and a semantic analysis module (320).

[0064] The pronunciation analysis module (310) analyzes pronunciation in the source language and the target language to identify personal information elements of the target language. The pronunciation analysis module (310) may include a pronunciation conversion module. The pronunciation conversion module can convert text representing personal information in the original text of the source language into phonetic symbols. The pronunciation conversion module can convert translated text of the target language into phonetic symbols. The pronunciation analysis module (310) may include a similarity calculation module. The similarity calculation module can calculate a first score representing similarity between phonetic symbols. Based on the first score, the pronunciation analysis module (310) can identify personal information elements within the target language that correspond to personal information of the source language.

[0065] The semantic analysis module (320) analyzes the semantics in the source language and the target language to identify personal information elements of the target language. The semantic analysis module (320) may be based on an artificial intelligence architecture (e.g., Transformer) that utilizes an attention mechanism. The attention module may include an embedding layer that converts the original text of the source language and the translated text of the target language into vectors. The attention module may calculate a second score, which is an attention score representing the degree of association between words in each text, by calculating the association between the original text of the source language and the translated text of the target language. Based on the second score, the semantic analysis module (320) may identify personal information elements within the target language that correspond to personal information of the source language.

[0066] The cross-language model (300) can detect personal information elements of the target language based on the first score and the second score, which are the analysis results of the pronunciation analysis module (310) and the semantic analysis module (320). For example, the cross-language model (300) can determine a common element as a personal information element among the first personal information element candidate determined by the first score and the second personal information element candidate determined by the second score. For example, the cross-language model (300) can determine the element having the highest overall score as a personal information element based on the sum of the first score and the second score. For example, the cross-language model (300) can determine a personal information element by applying a weight corresponding to at least one of the characteristics of the language and the characteristics of the personal information.

[0067] Meanwhile, the pronunciation analysis module (310) and semantic analysis module (320) illustrated in FIG. 3 may be configurations implemented by a processor included in an electronic device loading a program or instruction stored in the storage of the electronic device into memory and executing it. Also, each module illustrated in FIG. 3 is an example for convenience of explanation. For example, one module may be divided into multiple modules distinguished according to detailed functions. For example, a translation module that performs a translation function may be further included in the cross-language model (300). Alternatively, the functions of the modules described separately may be combined and implemented as a single module.

[0068] FIG. 4 is a diagram illustrating the operation of a pronunciation analysis module of a cross-language model according to one embodiment of the present disclosure.

[0069] Referring to FIG. 4, the pronunciation analysis module (400) may include a pronunciation conversion module (410) and a similarity calculation module (420).

[0070] In one embodiment, the pronunciation conversion module (410) can convert text of each language into pronunciation in each language. For example, the pronunciation conversion module (410) can convert text into International Phonetic Alphabet (IPA) notation. The conversion to IPA notation can be performed based on a standard IPA database corresponding to each language.

[0071] The pronunciation conversion module (410) can receive personal information detected in the original text of the source language and convert it into IPA notation. For example, if the source language is Korean, the personal information detected in the original text may be the Korean name 'Hong Gil-dong'. For the convenience of explanation, the personal information 'Hong Gil-dong' is illustrated as a single piece, but there may be multiple pieces of personal information detected in the original text. The pronunciation conversion module (410) converts the Korean text 'Hong Gil-dong' into the pronunciation in Korean. It can be converted into. In other words, the IPA notation of personal information (text) can be obtained.

[0072] The pronunciation conversion module (410) can receive text translated into a target language and convert it into an IPA notation. For example, if the target language is English, the translated text includes 'Hong Gil-Dong,' which is the English translation of 'Hong Gil-Dong' in the original text. The electronic device can translate the entire original text of the source language, and accordingly, the pronunciation conversion module (410) can obtain the IPA notation of the translated text. For example, 'Hong Gil-Dong' included in the translated text is converted into an IPA notation and the pronunciation in English This can be.

[0073] In one embodiment, the similarity calculation module (420) can calculate the similarity between pronunciations of different languages. For example, the similarity calculation module (420) can calculate the similarity between IPA notations of different languages.

[0074] The similarity calculation module (420) can calculate the similarity between the IPA notation of personal information output from the pronunciation conversion module (410) and the IPA notation of the translated text. The similarity calculation module (420) can calculate the similarity between IPA notations using various similarity calculation algorithms. For example, the similarity calculation module (420) may use similarity calculation algorithms such as cosine similarity, Euclidean distance, Jaccard similarity, and Levenstein distance, but is not limited thereto.

[0075] For example, the similarity calculation module (420) can calculate the similarity between the IPA notation of personal information and the IPA notation of translated text using cosine similarity. The similarity calculation module (420) can convert the IPA notation of personal information and the IPA notation of translated text into vectors, respectively, and calculate Cosine Similarity(IPA1, IPA2), which is a similarity score indicating how similar the directionality is between each vector. IPA1 represents the IPA notation of personal information in the source language (first language), and IPA2 represents the IPA notation of text translated into the target language (second language).

[0076] For example, the similarity calculation module (420) can calculate the similarity between the IPA notation of personal information and the IPA notation of translated text using Euclidean distance. The similarity calculation module (420) can convert the IPA notation of personal information and the IPA notation of translated text into vectors, respectively, and calculate a similarity score, Euclidean Distance (IPA1, IPA2), which represents the distance between each vector.

[0077] For example, the similarity calculation module (420) can calculate the similarity between the IPA notation of personal information and the IPA notation of translated text using Jaccard similarity. The similarity calculation module (420) can calculate the ratio of common elements (intersection) and total elements (union) between the sets of strings of the IPA notation of personal information and the set of strings of the IPA notation of translated text, and calculate the Jaccard Similarity (IPA1, IPA2), which is a similarity score indicating how much the symbols used in the two pronunciations overlap.

[0078] For example, the similarity calculation module (420) can calculate the similarity between the IPA notation of personal information and the IPA notation of translated text using the Levenshtein distance. The similarity calculation module (420) can calculate the Levenshtein Distance (IPA1, IPA2), which is a similarity score indicating how many editing operations are required to match the two strings, by measuring the editing distance between the string of the IPA notation of personal information and the string of the IPA notation of translated text.

[0079] In the aforementioned similarity calculation algorithms, cosine similarity and Jaccard similarity have values ​​between 0 and 1, and the closer the value is to 1, the more similar the two data are. Also, Euclidean distance and Levenstein distance indicate that the closer the value is to 0, the more similar the two data are.

[0080] In one embodiment, the pronunciation analysis module (400) can calculate a comprehensive similarity score by combining similarity scores using a plurality of similarity analysis algorithms. To combine similarity scores having different characteristics, the pronunciation analysis module (400) can normalize the Euclidean distance and the Levenshtein distance. For example, the normalized Euclidean distance may be 1 / (1+ Euclidean Distance) and the normalized Levenshtein distance may be 1 / (1+ Levenshtein Distance), and the normalized result may have a value between 0 and 1, and the closer the value is to 1, the more similar the two data are.

[0081] The pronunciation analysis module (400) can calculate a comprehensive similarity score by combining similarity scores or normalized similarity scores with a weighted average. The comprehensive similarity score can be calculated as follows.

[0082]

[0083] In the above equation, Score1 may represent cosine similarity, Score2 may represent Jaccard similarity, Score3 may represent normalized Euclidean distance, and Score4 may represent normalized Levenstein distance. w1, w2, w3, and w4 may be weights for each score, and the sum of the weights may be w1+w2+w3+w4=1.

[0084] The electronic device can use a pronunciation analysis module (400) to calculate a first score, which is a similarity score representing the similarity between the IPA notation of personal information and the IPA notation of translated text. In the example of FIG. 4, as a result of performing pronunciation analysis on the IPA notation of the personal information ‘Hong Gil-dong’, the similarity of the IPA notation of ‘Hong Gil-Dong’ among the translated texts may have a relatively high value, and the similarity of the IPA notation of the remaining text among the translated texts may have a relatively low value.

[0085] In one embodiment, the electronic device may apply weights to the similarity between the IPA notation of the translated text and the IPA notation of the personal information based on groups of phonetic symbols that group similar phonetic symbols. For example, the pronunciation analysis module (400) may generate groups of phonetic symbols that group phonetic symbols having phonetic similarity in the IPA notation of each language. The pronunciation analysis module (400) may assign high weights to phonetic symbols belonging to the same group in the IPA notations of two languages, considering them to have high similarity, and assign low weights to phonetic symbols belonging to different groups, considering them to have low similarity. The weights determined based on the phonetic symbol groups may be applied to the similarity between the IPA notation of the translated text and the IPA notation of the personal information. By applying weights using phonetic symbol groups, the electronic device may correct for subtle pronunciation differences existing between different languages. For example, even if the similarity between the IPA notation of the translated text and the IPA notation of personal information is calculated to be relatively low due to subtle pronunciation differences between the languages, if the phonetic symbols included in each IPA notation belong to corresponding phonetic symbol groups, weights can be applied to correct the similarity.

[0086] FIG. 5 is a diagram illustrating the operation of a semantic analysis module of a cross-language model according to one embodiment of the present disclosure.

[0087] Referring to FIG. 5, the semantic analysis module (500) may include an association calculation module (510). In describing FIG. 5, for convenience of explanation, the source language is Korean and the target language is English. Furthermore, the original text of the source language contains one or more pieces of personal information, and the association is calculated using the Korean name 'Hong Gil-dong' as an example among the one or more pieces of personal information.

[0088] In one embodiment, the association calculation module (510) can calculate the association between words in the original text and words in the translated text. The association calculation module (510) can receive the original text of the source language and the text translated into the target language as input and calculate the association between the two texts. The association calculation module (510) can use an attention mechanism (e.g., cross-attention) to calculate the association.

[0089] For example, the original text may be the Korean sentence "My name is Hong Gil-dong" and the translated text may be the English sentence "My name is Hong Gil-Dong". The semantic analysis module (500) can convert each sentence into an embedding vector. The embedding vector may utilize a pre-trained embedding model such as Word2Vec or BERT, but is not limited thereto. In the example of FIG. 5, the original text contains four words, "Je", "Name is", "Hong Gil-dong", and "is", so it is converted into four vectors, and the translated text contains five words, "My", "name", "is", "Hong", and "Gil-Dong", so it can be converted into five vectors.

[0090] The association calculation module (510) can calculate the association between two sentences using an attention mechanism. For example, the association calculation process using the attention mechanism calculates the similarity between a query and a key and normalizes it to obtain an attention score. Specifically, for example, word vectors of the translated text may correspond to the query, and word vectors of the original text may correspond to the key. In this case, similarity may be calculated for the query "My" and the keys "Je", "Im-eun", "Hong Gil-dong", and "Im-eun", respectively. Similarity calculation may use, for example, a dot product or cosine similarity, but is not limited thereto.

[0091] The association calculation module (510) can calculate an attention score representing the association between words as follows by applying a softmax function to the calculated similarity and normalizing it.

[0092]

[0093] The attention score matrix calculated by applying the example of Fig. 5 to the above equation can be expressed as follows.

[0094]

[0095] In the matrix above, represents the attention score between the i-th word (query) of the translated text and the j-th word (key) of the original text.

[0096] In the example of Fig. 5, the word representing personal information within the original text "My name is Hong Gil-dong" is "Hong Gil-dong," which is the third word in the original text. In the translated text "My name is Hong Gil-Dong," the words representing personal information are "Hong" (surname) and "Gil-Dong" (given name), which are the fourth and fifth words in the translated text. (Note: In the example of Fig. 5, since the format of a Korean name lists the surname first and the given name later, the English name translated from the Korean name was explained as an example where the surname is listed first and the given name later. However, depending on the format of the English name, the given name may be listed first and the surname later.)

[0097] The semantic analysis module (500) can identify words within the translated text that have a high degree of association with the word corresponding to personal information in the original text based on the attention score. For example, since personal information is the 3rd word of the original text, the semantic analysis module (500) within the attention score matrix The values ​​of can be identified. In this case, since the 3rd word of the original text and the 4th and 5th words of the translated text have a high correlation, Among the values ​​of The values ​​of can have relatively high values.

[0098] FIG. 6 is a diagram illustrating the operation of an electronic device according to one embodiment of the present disclosure to detect personal information by combining pronunciation analysis and semantic analysis.

[0099] If an electronic device relies solely on pronunciation analysis or solely on semantic analysis to detect cross-linguistic personal information, false positives may occur where items that are not actually personal information are identified as such. This arises due to the characteristics of cross-linguistics and personal information; therefore, the electronic device may utilize a combination of pronunciation and semantic analysis to prevent false positives.

[0100] For example, referring to the first block (610) which shows the result using only the pronunciation analysis module, the personal information to be detected may be the English name "Gina". In the process of detecting personal information in cross-language, if personal information is detected by analyzing pronunciation in Korean text, a false positive may occur because the word "Jina" in the sentence "Time passed, and it became summer again" is calculated to have a similar pronunciation to the word "Gina". However, in the Korean sentence, "Jina" is not the English name "Gina" written in Korean, but a word meaning "to pass." Nevertheless, such a false positive may occur when detecting personal information in cross-language based only on pronunciation.

[0101] For example, by referring to the second block (620) which represents the result using only the semantic analysis module, the semantic association between the Korean text "Hong Gil-Dong is my friend, but He is a really nice person" and the translation of the text "Hong Gil-Dong is my friend, but He is a really nice person" can be calculated. In the Korean text, the personal information is the Korean name "Hong Gil-Dong," and in the English text, "Hong Gil-Dong" and the pronoun "He" can be detected as words that have a semantic association with "Hong Gil-Dong." However, since the pronoun "He" is not personal information, this can be classified as a false positive.

[0102] To prevent false positives that may occur depending on the characteristics of cross-language personal information, the electronic device of the present disclosure combines pronunciation analysis and semantic analysis for the detection of cross-language personal information.

[0103] In one embodiment, the electronic device may calculate a first score (e.g., a similarity score) representing the similarity between personal information in the source language and sentences and words within the text translated into the target language, based on pronunciation analysis. Additionally, the electronic device may calculate a second score (e.g., an association score) representing the association between personal information in the source language and sentences and words within the text translated into the target language, based on semantic analysis. Based on the first score and the second score, the electronic device may identify candidates for personal information elements to be detected in the translated text of the target language and determine the final personal information elements among the candidates.

[0104] In one embodiment, the electronic device may determine weights to combine pronunciation analysis and semantic analysis. The weights correspond to characteristics of the language and / or personal information and may be learnable values. For example, pronunciation weights and semantic weights corresponding to characteristics of the language "Korean" may be pre-calculated, and pronunciation weights and semantic weights corresponding to each personal information characteristic, such as "name" and "address," may be pre-calculated. Based on the characteristics of the language and personal information, the electronic device may determine weights to combine pronunciation analysis and semantic analysis.

[0105] In one embodiment, the electronic device may apply weights that reflect the characteristics of the language. The weights that reflect the characteristics of the language may include pronunciation weights and semantic weights. For example, the first language (e.g., source language) and the second language (e.g., target language) being compared may have a statistically low pronunciation similarity. In this case, the electronic device may calculate a weighted sum of the first score and the second score by applying pronunciation weights and semantic weights so that the second score based on semantic analysis is reflected more than the first score based on pronunciation analysis. Similarly, if the pronunciation similarity between the first language and the second language is high, the electronic device may apply pronunciation weights and semantic weights so that the first score based on pronunciation analysis is reflected more. The pronunciation weights and semantic weights that reflect the characteristics of the language may be pre-calculated for each language.

[0106] In one embodiment, the electronic device may apply weights that reflect personal information characteristics. The weights that reflect personal information characteristics may include pronunciation weights and semantic weights. For example, in the case of personal information such as names, Korean names are listed in the order of surname-given name, and English names are listed in the order of given name-surname. Or, in the case of personal information such as addresses, Korean addresses are listed in the order of larger scope to smaller scope (e.g., country-city-building), and English addresses are listed in the order of smaller scope to larger scope (e.g., building-city-country). Since the degree to which pronunciation similarity or semantic similarity plays an important role varies for each of these personal information characteristics, the electronic device may calculate a weighted sum by applying pronunciation weights to a first score or applying semantic weights to a second score according to the personal information characteristics. Based on the weighted sum of the first score and the second score to which weights are applied, the electronic device may detect personal information elements corresponding to the personal information of the original text in the translated text.

[0107] In one embodiment, the electronic device may use a combination of weights based on language characteristics and weights based on personal information characteristics. For example, the electronic device may calculate a weighted sum of a first score and a second score by using a first pronunciation weight and a first semantic weight, which are weights based on language characteristics, and a second pronunciation weight and a second semantic weight, which are weights based on personal information characteristics.

[0108] FIG. 7 is a diagram illustrating an example of an operation in which an electronic device according to one embodiment of the present disclosure detects personal information across languages.

[0109] In one embodiment, the electronic device may perform a security check on the conversation between the user and the chatbot. For example, the chatbot system may be composed of a client device (700) and a server (710). The client device (700) may input a question to the server (710), and the server (710) may provide an answer to the client device (700). The server (710) may be a server that operates a language model (e.g., a large-scale language model) for providing chatbot services.

[0110] Chatbot services provide user convenience by answering user questions. However, the content of chatbot responses must not include sensitive personal information. To inspect and improve the security functions of chatbot services, it is necessary to detect whether personal information is included in the chatbot's responses. Additionally, since chatbot services provide responses in the language corresponding to the language used by the user, it is necessary to detect whether personal information is included in the chatbot's responses across multiple languages. Electronic devices can use a cross-language model to check whether personal information is included in chatbot responses across various languages ​​using a single model, without modifying the model as the language changes.

[0111] In one embodiment, the electronic device may collect conversation content between a client device (700) and a server (710). The electronic device may receive conversation content from the client device (700) or receive conversation content between a user and a chatbot from the server (710). Alternatively, the electronic device may acquire conversation content between a user and a chatbot while operating as the client device (700) or the server (710).

[0112] In one embodiment, an electronic device can detect personal information in the content of a conversation between a user and a chatbot. The electronic device can detect personal information in the content of the conversation using a cross-language model. The cross-language model can perform pronunciation analysis and semantic analysis tasks to detect personal information in the text of a target language based on personal information elements corresponding to the source language. Since the processes of preliminary tasks (e.g., personal information detection, translation, etc.) and analysis tasks (e.g., pronunciation analysis and semantic analysis) required for the analysis tasks of the cross-language model have been described in previous drawings, a repetitive explanation is omitted.

[0113] To explain how an electronic device detects cross-language personal information in the conversation content of a chatbot service, the detection of personal information in the source language, "Hong Gil-dong" (Korean name) and "56 Seongchon-gil, Seocho-gu" (Korean address), is described as an example.

[0114] For example, if a question in Korean is entered into the chatbot, the chatbot may output a Korean answer sentence "Hong Gil-Dong is living at Seocho-gu Seongchon-gil 56." In this case, the electronic device may use a cross-language model to detect the personal information "Hong Gil-Dong" and "Seocho-gu Seongchon-gil 56" in Korean from the chatbot's answer, and perform a security check to determine whether the chatbot outputs sensitive personal information.

[0115] In addition, the electronic device can use a cross-language model to check whether the chatbot's answer contains sensitive personal information when a question in a different language is input. For example, if a question in English is input to the chatbot, the chatbot may output the English answer sentence "Hong Gil-Dong is living at Seocho-gu Seongchon-gil 56". In this case, the electronic device can use the cross-language model without changing the model to detect the English personal information "Hong Gil-Dong" and "Seocho-gu Seongchon-gil 56" in the chatbot's answer, and perform a security check to determine whether the chatbot outputs sensitive personal information.

[0116] The electronic device can perform chatbot security checks for various languages ​​and update the chatbot's security functions.

[0117] FIG. 8 is a diagram illustrating an example of an operation in which an electronic device according to one embodiment of the present disclosure detects personal information across languages.

[0118] In one embodiment, the electronic device may perform a security check on a document or a database. For example, the electronic device may detect whether the document (800) (or the database containing text) contains personal information that is sensitive information. The document (800) may be a document in a first language (e.g., Korean), a document in a second language (e.g., English), or a document in which the first and second languages ​​are mixed. Alternatively, the document (800) may contain two or more languages.

[0119] The electronic device can detect personal information within a document (800) using a cross-language model. Since the cross-language model detects personal information using pronunciation analysis and semantic analysis, it can detect personal information regardless of what language the personal information is written in.

[0120] For example, in the Korean sentence "Hong Gil-Dong is living at Seocho-gu Seongchon-gil 56," {Korean-Korean Name: Hong Gil-Dong} and {Korean-Korean Address: Seocho-gu Seongchon-gil 56} can be detected. Also, for example, in the Korean sentence "James is taking medicine because of diabetes," {Korean-American Name: James} can be detected. This means that even if personal information in English names is written in Korean, the cross-language model can detect it as personal information.

[0121] For example, in the English sentence "Hong Gil-Dong is living at Seocho-gu Seongchon-gil 56", {English-Korean Name: Hong Gil-Dong} and {English-Korean Address: Seocho-gu Seongchon-gil 56} can be detected. This means that even if personal information such as a Korean name is written in English, or a Korean address is written in English, the cross-language model can detect this as personal information. Also, for example, in the English sentence "James is taking medicine because of diabetes", {English-American Name: James} can be detected.

[0122] In other words, the electronic device can detect personal information for different languages ​​using a single cross-language model without changing to a language model corresponding to each language whenever the language changes.

[0123] In one embodiment, in addition to detecting cross-language personal information within a single document, the electronic device may also perform security checks on original and translated documents. For example, the electronic device may perform security checks to detect personal information on original documents and their translations using a single cross-language model.

[0124] In one embodiment, the electronic device may determine the level of protection of a document based on personal information elements detected in the document (800) (or personal information elements detected in a database). For example, the electronic device may set access rights to the document based on the number of detected personal information elements. Alternatively, the electronic device may protect (e.g., mask) at least some of the personal information elements based on the type of the detected personal information elements.

[0125] FIG. 9 is a flowchart illustrating the operation of an electronic device according to one embodiment of the present disclosure applying settings for personal information detection.

[0126] In operation S910, the electronic device may obtain user input selecting a target language. Operation S910 may be performed after operation S220 of FIG. 2 has been performed. For example, if the source language is a first language (e.g., Korean), the electronic device may obtain user input selecting a second language (e.g., English) as the target language, which is a different language.

[0127] In operation S920, the electronic device can determine whether to perform pronunciation analysis based on identification information of the source language and the target language.

[0128] Since differences in pronunciation systems and phonological rules exist among languages, there may be languages ​​suitable for pronunciation analysis and languages ​​unsuitable for it. For example, languages ​​with similar pronunciation systems or common rules may be suitable for pronunciation comparison, whereas languages ​​with pronunciation rules that differ somewhat from others (e.g., languages ​​where meanings vary depending on tone, or languages ​​containing voiceless consonants) may be unsuitable for such comparison. The electronic device can determine whether to perform pronunciation analysis based on predefined information regarding whether to perform pronunciation analysis between the source language and the target language.

[0129] In one embodiment, operation S920 may be omitted. In other words, operation S930 may be performed immediately after operation S910 is performed. In this case, it means that pronunciation analysis is always performed.

[0130] In operation S930, the electronic device can apply settings for performing pronunciation analysis and semantic analysis.

[0131] Once the source language and target language are determined, the electronic device can load settings for performing pronunciation analysis and semantic analysis between the languages. For example, the electronic device can load IPA information of the source language for pronunciation conversion of the source language and IPA information of the target language for pronunciation conversion of the target language, and apply them as settings for pronunciation analysis. Additionally, for example, the electronic device can load a translation model for translation between the source language and the target language and apply it as settings for semantic analysis. In one embodiment, translation between the source language and the target language may include loading a translation model that translates into an intermediate language, which is a third language.

[0132] After operation S930 is performed, operation S230 of FIG. 2 may be performed. Since subsequent operations have been described in the description of FIG. 2, a repetitive description is omitted.

[0133] FIG. 10 is a flowchart illustrating the operation of an electronic device detecting personal information according to one embodiment of the present disclosure.

[0134] In operation S1010, the electronic device may acquire text in a first language. The text in the first language may refer to the text on which the electronic device performs a personal information detection operation. The electronic device may acquire the text based on user input, acquire the text from a stored text file, or acquire the text from an external device. When the text in the first language is acquired, the electronic device may initiate operations to detect one or more personal information elements within the text in the first language. In this case, the electronic device may detect that personal information originally in a second language is written in the first language due to reasons such as translation.

[0135] In operation S1020, the electronic device may obtain personal information in a second language. The second language may refer to a language different from the first language. For example, if the first language is Korean, the second language may be English. Personal information in the second language may include, for example, name, gender, date of birth, address, email address, phone number, social media account, resident registration number, passport number, driver's license number, etc., but is not limited thereto.

[0136] In operation S1030, the electronic device can calculate a first score based on pronunciation analysis.

[0137] In one embodiment, the electronic device can convert text of a first language into a phonetic representation. For example, the electronic device can convert text of the first language into an IPA notation corresponding to the first language. Additionally, the electronic device can convert personal information of a second language into a phonetic representation. For example, the electronic device can convert personal information of the second language into an IPA notation corresponding to the second language. The electronic device can calculate a first score indicating the degree of similarity by calculating the similarity between the IPA notation of the text of the first language and the IPA notation of the personal information of the second language.

[0138] In one embodiment, the electronic device may determine candidates for personal information elements based on a first score. For example, the electronic device may determine candidates for first personal information elements within a text of a first language based on whether the first score exceeds a first threshold. Or, for example, the electronic device may detect candidates for first personal information elements within a text of a first language based on the top N words with high first scores.

[0139] In operation S1040, the electronic device can calculate a second score based on semantic analysis.

[0140] For example, the electronic device can translate the entire text of the first language into the second language. The electronic device can calculate a second score representing the degree of association between the text translated into the second language and the personal information of the second language. The calculation of the association may be based on an attention score obtained using an attention mechanism, but is not limited thereto. The electronic device can determine candidates for second personal information elements within the text of the first language based on whether the second score exceeds a second threshold. Alternatively, for example, the electronic device can detect candidates for second personal information elements within the text of the first language based on the top M words with high second scores.

[0141] In one embodiment, for example, the electronic device may translate only the first personal information element candidates, which are personal information element candidates determined by the first score, into the second language. The electronic device may calculate a second score representing the degree of association between the first personal information element candidates translated into the second language and the personal information in the second language. The electronic device may reduce the amount of computation by translating only the first personal information element candidates detected in the text of the first language, rather than translating the entire text of the first language into the second language. The degree of association may be an attention score obtained using an attention mechanism, but is not limited thereto.

[0142] In operation S1050, the electronic device can detect personal information elements written in the first language that correspond to personal information in the second language in the text of the first language, based on the first score and the second score.

[0143] The electronic device can determine a final personal information element from among the candidates for identified personal information elements based on the first score and the second score. The final personal information element may be one or more and may include cross-language personal information. Cross-language personal information refers to personal information written in the first language that corresponds to personal information in the second language. For example, personal information written in the first language corresponding to the English name "James," which is personal information in the second language, may be the Korean name "James." In other words, the English name "James" written in Korean can be detected within text written in the first language.

[0144] In one embodiment, the electronic device may determine a common element as a personal information element among a first personal information element candidate determined by a first score and a second personal information element candidate determined by a second score. Alternatively, the electronic device may determine the element having the highest overall score as a personal information element based on the sum of the first score and the second score.

[0145] In one embodiment, the electronic device may calculate a second score by translating only the first personal information element candidates determined by the first score, and then determine an element whose second score is greater than or equal to a threshold value as a personal information element. Alternatively, the electronic device may calculate a second score by translating only the first personal information element candidates determined by the first score, and then determine an element with the highest second score as a personal information element.

[0146] In one embodiment, the electronic device may apply weights based on the characteristics of language and personal information. The electronic device may apply weights to a first score or a second score. Since specific operations related to the application of weights have been described in the description of the preceding drawings, a repetitive description is omitted.

[0147] FIG. 11 is a block diagram illustrating the configuration of an electronic device according to one embodiment of the present disclosure.

[0148] In one embodiment, the electronic device (1000) may include a communication interface (1100), a memory (1200), a processor (1300), and a display (1400).

[0149] The communication interface (1100) can perform data communication with other electronic devices under the control of the processor (1300).

[0150] The communication interface (1100) can perform data communication between an electronic device (1000) and another electronic device by using at least one of data communication methods including, for example, wired LAN (e.g., Ethernet), wireless LAN (e.g., Wi-Fi), cellular network (e.g., 4G, 5G, etc.), Bluetooth, BLE (Bluetooth Low Energy), ZigBee, infrared communication (IrDA, infrared Data Association), NFC (Near Field Communication), RF communication, and various other types of known wireless / wired communication technologies. The communication interface (1100) may include a communication circuit designed to use the aforementioned communication methods.

[0151] An electronic device (1000) can transmit and receive data for detecting cross-language personal information to and from another electronic device using a communication interface (1100). For example, the electronic device (1000) can transmit and receive input data (e.g., text) of a cross-language model and / or output data (e.g., personal information detection results) of a cross-language model to and from another electronic device, and can receive a model (e.g., a cross-language model, a translation model, etc.) for performing a personal information detection task from another electronic device.

[0152] Memory (1200) may include various types of memory. Memory (1200) may include a main memory that stores data currently being processed in an electronic device (1000). For example, the main memory may include volatile memory such as RAM (Random Access Memory) or SRAM (Static Random Access Memory), but is not limited thereto. Memory (1200) may include a secondary memory that permanently stores large amounts of data (e.g., programs, system files, etc.). For example, the secondary memory may include non-volatile memory including at least one of a hard disk drive (HDD), a solid-state drive (SSD), an optical drive (e.g., CD), a flash drive, ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), and PROM (Programmable Read-Only Memory), but is not limited thereto.

[0153] The memory (1200) may store one or more instructions and one or more programs that enable the electronic device (1000) to operate to detect and provide cross-language personal information. For example, the memory (1200) may store instructions and programs for implementing the functions of the pronunciation analysis module (1210) and the semantic analysis module (1220). The modules stored in the memory (1200) may be examples for convenience of explanation, and the present disclosure is not limited to such examples. Other modules may be added to implement the aforementioned embodiments, and some modules may be omitted. Additionally, one module may be divided into a plurality of modules distinguished by detailed functions, and some of the aforementioned modules may be combined to be implemented as a single module.

[0154] The processor (1300) can control the overall operations of the electronic device (1000). The processor (1300) may include a processing circuit. For example, the processor (1300) can control the overall operations of the electronic device (1000) for detecting and processing cross-language personal information by executing one or more instructions of a program stored in memory (1200). There may be one or more processors (1300).

[0155] The processor (1300) may be composed of at least one of, for example, a Central Processing Unit (CPU), a Microprocessor, a Graphic Processing Unit (GPU), ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), an Application Processor (AP), a Neural Processing Unit (NPU), or an AI-dedicated processor designed with a hardware structure specialized for processing AI models, but is not limited thereto.

[0156] The processor (1300) can execute a pronunciation analysis module (1210) to perform a pronunciation analysis task for detecting cross-language personal information. The pronunciation analysis module (1210) can analyze pronunciations between different languages ​​and calculate a similarity score indicating the pronunciation similarity between sentences and words of each language. Since the description of the operations related to the pronunciation analysis module (1210) has already been described previously in the description of the drawings, a repetitive description is omitted.

[0157] The processor (1300) can execute a semantic analysis module (1220) to perform semantic analysis for cross-language personal information detection. The semantic analysis module (1220) can analyze the semantics between different languages ​​and calculate an association score indicating the semantic association between sentences and words of each language. Since the descriptions related to the operations of the semantic analysis module (1220) have already been described in the description of the previous drawings, a repetitive description is omitted.

[0158] In one embodiment, there may be one or more processors (1300). If there is one or more processors (1300), the operations of the present disclosure may be performed by one or more processors individually or collectively by executing instructions and / or programs stored in memory (1200). If the method according to one embodiment of the present disclosure includes a plurality of operations, the plurality of operations may be performed by one processor (1300) or by a plurality of processors (1300).

[0159] For example, when the first, second, and third operations are performed by a method according to one embodiment, the first, second, and third operations may all be performed by a first processor, or some of the first to third operations may be performed by a first processor (e.g., a general-purpose processor) and the remaining operations may be performed by a second processor (e.g., an AI-dedicated processor). Here, operations for training / inference of an AI model may be performed by an AI-dedicated processor, which is an example of a second processor. However, the embodiments of the present disclosure are not limited thereto.

[0160] One or more processors according to the present disclosure may be implemented as a single-core processor or as a multi-core processor. When a method according to one embodiment of the present disclosure includes a plurality of operations, the plurality of operations may be performed by a single core or by a plurality of cores included in one or more processors.

[0161] The display (1400) can output a video signal to the screen of the electronic device (1000) under the control of the processor (1300). The display (1400) can output a video signal processed during the process of the electronic device (1000) detecting cross-language personal information to the screen. For example, the display (1400) can display an input field for text input, a text list for selecting an input text file, text search results, selected text, etc., and can display the results of detecting cross-language personal information within the text. The display (1400) may include a touch panel. The touch panel may include one or more touch sensors that detect touch input. In one embodiment, user input related to the cross-language personal information detection task may be obtained through the touch panel.

[0162] FIG. 12 is a block diagram illustrating the configuration of a server according to one embodiment of the present disclosure.

[0163] In one embodiment, the server (2000) may include a communication interface (2100), a memory (2200), and a processor (2300). The operations of the electronic device (1000) described in the preceding drawings may be performed by the server (2000).

[0164] The communication interface (2100), memory (2200), and processor (2300) of the server (2000) of FIG. 12 can correspond, respectively, to the basic functions described in the communication interface (1100), memory (1200), and processor (1300) of the electronic device (1000) of FIG. 11. Therefore, for brevity, repetitive descriptions are omitted. The memory (2200) can store commands and programs for implementing the functions of the pronunciation analysis module (2210) and the semantic analysis module (2220).

[0165] The server (2000) may be a computing device composed of hardware elements having higher performance specifications than the electronic device (1000) so as to be able to process complex operations and tasks for processing large-scale data generated during the operation of a cross-language model. Accordingly, each component of the server (2000) may have similar functions to each component of the electronic device (1000), but may have higher specifications in terms of performance (e.g., amount of computation, computation speed, etc.).

[0166] The server (2000) receives a request for a personal information detection task using a cross-language model from a client device (e.g., a user device), performs a cross-language personal information detection task based on pronunciation analysis and semantic analysis, and can provide the detected personal information results to the client device as a response.

[0167] The present disclosure relates to an electronic device and method for detecting personal information elements within text using a single cross-language model capable of detecting personal information in several different languages. The technical problems to be solved by the present disclosure are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description in this specification.

[0168] According to one aspect of the present disclosure, a method for detecting personal information performed by an electronic device may be provided.

[0169] The above method may include the step of obtaining a first text of a first language.

[0170] The above method may include a step of detecting personal information in the first text.

[0171] The above method may include the step of translating the first text into a second language to obtain a second text.

[0172] The above method may include the step of identifying a first score based on the pronunciation analysis of the above personal information and the above second text.

[0173] The above method may include the step of identifying a second score based on semantic analysis of the above personal information and the above second text.

[0174] The above method may include the step of detecting personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

[0175] The step of identifying the first score may include converting the personal information and the second text into International Phonetic Alphabet (IPA) notation.

[0176] The step of identifying the first score may include identifying the first score representing the similarity between the IPA notation of the second text and the IPA notation of the personal information.

[0177] The step of identifying the first score may include applying weights to the similarity between the IPA notation of the second text and the IPA notation of the personal information based on a group of phonetic symbols grouped with similar phonetic symbols.

[0178] The step of identifying the second score may include the step of identifying the second score representing the degree of association between the first text and the second text by applying an attention mechanism.

[0179] The step of obtaining the second text may include the step of translating the first text into a third language, which is an intermediate language.

[0180] The step of obtaining the second text may include the step of obtaining the second text by translating the third text translated into the third language into the second language.

[0181] The above method may include the step of applying a weight corresponding to at least one of the language characteristic or the personal information characteristic to at least one of the first score or the second score.

[0182] The above method may include the step of determining the priority between the pronunciation analysis and the semantic analysis by comparing at least one of the first score or the second score with a threshold value.

[0183] The above method may include the step of obtaining user input selecting the second language.

[0184] The above method may include the step of applying settings for performing pronunciation analysis and semantic analysis based on identification information of the first language and the second language.

[0185] The above method may include a step of determining whether to perform the pronunciation analysis based on identification information of the first language and the second language.

[0186] The above method may include the step of determining the protection level of a document containing the second text based on personal information elements detected from the second text.

[0187] According to one aspect of the present disclosure, an electronic device for detecting personal information may be provided.

[0188] The electronic device may include a communication interface; at least one processor; and a memory for storing instructions.

[0189] By executing the above instructions by the at least one processor, the electronic device can obtain a first text of a first language.

[0190] By executing the above instructions by the at least one processor, the electronic device can detect personal information in the first text.

[0191] By executing the above instructions by the at least one processor, the electronic device can obtain a second text by translating the first text into a second language.

[0192] By executing the above instructions by the at least one processor, the electronic device can identify a first score based on the pronunciation analysis of the personal information and the second text.

[0193] By executing the above instructions by the at least one processor, the electronic device can identify a second score based on semantic analysis of the personal information and the second text.

[0194] By executing the above instructions by the at least one processor, the electronic device can detect personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

[0195] By executing the above instructions by the at least one processor, the electronic device can convert the personal information and the second text into International Phonetic Alphabet (IPA) notation.

[0196] By executing the above instructions by the at least one processor, the electronic device can identify the first score representing the similarity between the IPA notation of the second text and the IPA notation of the personal information.

[0197] By executing the above instructions by the at least one processor, the electronic device can apply weights to the similarity between the IPA notation of the second text and the IPA notation of the personal information based on a group of phonetic symbols grouping similar phonetic symbols.

[0198] By executing the above instructions by the at least one processor, the electronic device can identify the second score representing the degree of association between the first text and the second text by applying an attention mechanism.

[0199] By executing the above instructions by the at least one processor, the electronic device can translate the first text into a third language which is an intermediate language.

[0200] By executing the above instructions by the at least one processor, the electronic device can obtain the second text by translating the third text translated into the third language into the second language.

[0201] By executing the above instructions by the at least one processor, the electronic device can apply a weight corresponding to at least one of the language characteristic or the personal information characteristic to at least one of the first score or the second score.

[0202] By executing the above instructions by the at least one processor, the electronic device can determine the priority between the pronunciation analysis and the semantic analysis by comparing at least one of the first score or the second score with a threshold value.

[0203] By executing the above instructions by the at least one processor, the electronic device can obtain user input selecting the second language.

[0204] By executing the above instructions by the at least one processor, the electronic device can apply settings for the pronunciation analysis and the semantic analysis based on identification information of the first language and the second language.

[0205] By executing the above instructions by the at least one processor, the electronic device can determine whether to perform the pronunciation analysis based on identification information of the first language and the second language.

[0206] By executing the above instructions by the at least one processor, the electronic device can determine the level of protection of a document containing the second text based on personal information elements detected from the second text.

[0207] Embodiments of the present disclosure may also be implemented in the form of a recording medium containing computer-executable instructions, such as program modules executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, and both removable and non-removable media. Additionally, a computer-readable medium may include computer storage media and communication media. A computer storage medium includes both volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data. A communication medium may include other data of modulated data signals, such as computer-readable instructions, data structures, or program modules.

[0208] Additionally, computer-readable storage media may be provided in the form of non-transitory storage media. Here, 'non-transitory storage media' simply means that it is a tangible device and does not contain a signal (e.g., electromagnetic waves), and this term does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily. For example, 'non-transitory storage media' may include a buffer in which data is stored temporarily.

[0209] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product (e.g., a downloadable app) may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0210] The foregoing description of the present disclosure is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present disclosure. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0211] The scope of the present disclosure is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present disclosure.

Claims

1. A method for detecting personal information performed by an electronic device, Step of obtaining a first text of a first language; A step of detecting personal information from the first text above; A step of obtaining a second text by translating the first text into a second language; A step of identifying a first score based on pronunciation analysis of the above personal information and the above second text; A step of identifying a second score based on semantic analysis of the above personal information and the above second text; and A method comprising the step of detecting personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

2. In Paragraph 1, The step of identifying the first score above is, A step of converting the above personal information and the above second text into International Phonetic Alphabet (IPA) notation; and A method comprising the step of identifying the first score representing the similarity between the IPA notation of the second text and the IPA notation of the personal information.

3. In Paragraph 2, The step of identifying the first score above is, A method comprising the step of applying weights to the similarity between the IPA notation of the second text and the IPA notation of the personal information based on a group of phonetic symbols grouped with similar phonetic symbols.

4. In Paragraph 1, The step of identifying the second score above is, A method comprising the step of identifying a second score representing the degree of association between the first text and the second text by applying an attention mechanism.

5. In Paragraph 4, The step of obtaining the above second text is, A step of translating the above first text into a third language which is an intermediate language; and A method comprising the step of translating a third text translated into the third language into the second language to obtain the second text.

6. In Paragraph 1, The above method is, A method further comprising the step of applying a weight corresponding to at least one of a language characteristic or a personal information characteristic to at least one of the first score or the second score.

7. In Paragraph 1, The above method is, A method further comprising the step of determining a priority between the pronunciation analysis and the semantic analysis by comparing at least one of the first score or the second score with a threshold value.

8. In electronic devices, Communication interface; At least one processor; and It includes memory for storing instructions, By executing the above instructions by the at least one processor, the electronic device, Obtain the first text of the first language, and Detecting personal information from the above first text, and The above first text is translated into a second language to obtain a second text, and Identifying a first score based on the pronunciation analysis of the above personal information and the above second text, and Identifying a second score based on semantic analysis of the above personal information and the above second text, and An electronic device that detects personal information elements corresponding to the personal information of the first text in the second text based on the first score and the second score.

9. In Paragraph 8, By executing the above instructions by the at least one processor, the electronic device, Convert the above personal information and the above second text into International Phonetic Alphabet (IPA) notation, and An electronic device for identifying the first score indicating the similarity between the IPA notation of the second text and the IPA notation of the personal information.

10. In Paragraph 9, By executing the above instructions by the at least one processor, the electronic device, An electronic device that applies weights to the similarity between the IPA notation of the second text and the IPA notation of the personal information based on a group of phonetic symbols that groups similar phonetic symbols.

11. In Paragraph 8, By executing the above instructions by the at least one processor, the electronic device, An electronic device that identifies the second score representing the degree of association between the first text and the second text by applying an attention mechanism.

12. In Paragraph 8, By executing the above instructions by the at least one processor, the electronic device, Translate the above first text into a third language, which is an intermediate language, and An electronic device that obtains the second text by translating the third text translated into the third language into the second language.

13. In Paragraph 8, By executing the above instructions by the at least one processor, the electronic device, An electronic device that applies a weight corresponding to at least one of a language characteristic or a personal information characteristic to at least one of the first score or the second score.

14. In Paragraph 8, By executing the above instructions by the at least one processor, the electronic device, An electronic device that determines the priority between the pronunciation analysis and the semantic analysis by comparing at least one of the first score or the second score with a threshold value.

15. A computer-readable recording medium having a program for executing the method of any one of paragraphs 1 through 7 on a computer.

Images