Translation detection method and related equipment

Abstract

The embodiment of the present application provides a translation detection method and related equipment, the method comprising: obtaining data to be detected, the data to be detected comprising a source language text in a user interface and a corresponding translation; back-translating the translation to obtain a back-translation result; performing word segmentation processing on the back-translation result and the source language text respectively to obtain a first word segmentation and a second word segmentation; determining a first abnormal translation word segmentation in the second word segmentation based on the first word segmentation; screening the first abnormal translation word segmentation based on a translation necessity evaluation rule, and outputting the screened second abnormal translation word segmentation, the translation necessity evaluation rule comprising at least one of a part-of-speech rule, a scene association rule, and a target language style rule. Using the above method, by performing word segmentation and comparison on the source language text and the back-translation result, and using the translation necessity evaluation rule to evaluate the candidate abnormal translation word segmentations, it is possible to accurately screen out the missed / wrongly translated word segmentations in the source language text of the user interface.

Description

Technical Field

[0001] The present application belongs to the field of computer technology, relates to translation technology, and in particular to a translation detection method and related equipment. Background Art

[0002] In order to meet multilingual needs, user devices usually set up texts in multiple languages ​​for the same term before leaving the factory, so that texts in different languages ​​can be displayed according to the needs of different users. When translating terms in the user interface into multiple languages, if there are omissions or mistranslations, it may cause users to be unable to accurately understand the interface functions, affecting the normal use of the device or application, further affecting the user experience and reducing the user's favorability towards the product. Since user devices contain a large number of user interfaces, resulting in a large number of interface terms, manual multilingual translation review of interface terms is prone to omissions and inefficient.

[0003] In the translation of interface terms, the phenomenon of elision is relatively common due to the following three considerations: 1. The graphics in the user interface have a scene prompt function, and the words that are repeated in the graphic expression can be omitted when translating interface terms; 2. Due to the limited display area of ​​the user interface, it is easy to have incomplete text display problems. When translating interface terms, people tend to use more concise expressions; 3. Different languages ​​have differences in vocabulary, grammar and expression style. That is, when translating interface terms, elision strategies may be frequently used, and elision is not a problem of missing or wrong translation. Existing translation quality detection tools are prone to false positives when detecting the translation of interface terms. Summary of the invention

[0004] In view of the above, it is necessary to provide a translation detection method, which can improve the problem that existing translation quality detection tools are prone to false positives when performing translation detection of interface terms, and improve the detection accuracy of interface terminology translation.

[0005] In a first aspect, the present application provides a translation detection method, comprising: obtaining data to be detected, the data to be detected comprising a source language text in a user interface and a translation corresponding to the source language text; back-translating the translation to obtain a back-translation result corresponding to the translation; performing word segmentation processing on the back-translation result to obtain a first word segmentation, and performing word segmentation processing on the source language text to obtain a second word segmentation; determining a first translation abnormal word segmentation in the second word segmentation based on the first word segmentation; screening the first translation abnormal word segmentation based on a translation necessity assessment rule, and outputting the screened second translation abnormal word segmentation and the source language text corresponding to the second translation abnormal word segmentation, wherein the translation necessity assessment rule comprises at least one of a part-of-speech rule, a scene association rule, and a target language style rule.

[0006] By adopting the above technical scheme, a translation necessity assessment rule for word segmentation translation evaluation is constructed based on the characteristics of the user interface terminology. By back-translating the translation and comparing the back-translated result with the source language text (word segmentation processing and comparison), suspected abnormal translation segmentations are determined, and then the abnormal translation segmentations are screened and output based on the translation necessity assessment rule. This can achieve accurate screening of missed / wrongly translated segmentations in the source language text of the user interface.

[0007] In a possible implementation, the data to be detected also includes user interface information of the source language text, the user interface information includes the theme of the user interface and the name of the application to which the user interface belongs, and the scene association rule performs word segmentation and screening based on the user interface information.

[0008] By adopting the above technical solution, since the graphics in the user interface have the function of scene prompting, words that are repeated in the graphic expression are usually omitted when translating interface terms. By obtaining the user interface information of the source language text, it is possible to subsequently determine based on the scene association rules whether the omitted word segment is omitted because it overlaps with the user interface scene theme or application name, which can effectively reduce the false alarm of omitted interface terms.

[0009] In a possible implementation, back-translating the translation to obtain a back-translation result corresponding to the translation includes: back-translating the translation to obtain a first back-translation result; segmenting the translation to obtain third segmentations, and searching a bilingual dictionary and / or a bilingual term list to obtain a back-translation word corresponding to each third segmentation; and obtaining a back-translation result corresponding to the translation based on the first back-translation result and the back-translation word, wherein the back-translation result corresponding to the translation includes one or more back-translation texts corresponding to the translation.

[0010] By adopting the above technical scheme, since the back-translation of the translation usually adopts a machine translation model, and the user interface terms are mostly short sentences, most words can obtain relatively accurate translations through dictionaries and glossaries. By setting up a bilingual dictionary and a bilingual glossary to back-translate the word segments of the translation, a back-translated word set is obtained, and then a second back-translation result is constructed based on the first back-translation result and the back-translation word set, which can supplement the first back-translation result and reduce the back-translation accuracy loss that may be caused by back-translating the translation using a machine translation model.

[0011] In a possible implementation, after obtaining the back-translation result corresponding to the translation, the method further includes: calculating the similarity between the source language text and the corresponding back-translation result, and adding the source language text suspected of abnormal translation and the corresponding back-translation result to the text set based on the calculated similarity.

[0012] By adopting the above technical scheme, the similarity between the source language text and the corresponding back-translation result can be text similarity or semantic familiarity, and the similarity between the back-translation result and the corresponding source language text can be used to characterize the translation accuracy of the translation. By calculating the similarity between the source language text and the corresponding back-translation result, and screening the source language text suspected of having translation anomalies based on the calculated similarity, for example, filtering out the source language text with a similarity greater than a threshold, the source language text that is considered to have no translation anomalies is filtered out, and only the source language text with a similarity lower than the threshold and the corresponding back-translation result are added to the text set for subsequent processing, which can reduce the amount of data processing, save computing resources, and shorten the detection path of these texts without anomalies.

[0013] In a possible implementation, performing word segmentation processing on the back-translation result to obtain the first word segmentation, and performing word segmentation processing on the source language text to obtain the second word segmentation, includes: performing word segmentation processing on the back-translation result in the text set to obtain the first word segmentation, and performing word segmentation processing on the source language text in the text set corresponding to the back-translation result to obtain the second word segmentation.

[0014] By adopting the above technical solution, in the process of determining the candidate abnormal word segmentations in the source language text by word segmentation processing and comparison, only the source language text and the back-translation results in the text set are segmented, which can reduce the number of texts that need to be segmented, save computing resources, and improve the efficiency of translation detection.

[0015] In a possible implementation, determining a first abnormally translated word segment in a second word segment based on a first word segment includes: performing word frequency statistics on the first word segment to obtain a first word frequency result; performing word frequency statistics on the second word segment to obtain a second word frequency result; and determining the first abnormally translated word segment in the second word segment based on the first word frequency result and the second word frequency result.

[0016] By adopting the above technical solution, the word frequency statistics are respectively performed on the source language text and the word frequency results of the back-translation results, and the word frequency results of the two are compared, so that the suspected abnormal translation word segmentation in the source language text can be accurately determined. For example, if the word segmentation of the source language text does not appear in the back-translation result, or the word frequency number of the word segmentation is different from the word frequency number of the word segmentation in the back-translation result, it is determined to be a suspected abnormal translation word segmentation.

[0017] In a possible implementation, determining a first abnormally translated word segment in the second word segment based on the first word frequency result and the second word frequency result includes: determining the first abnormally translated word segment in the second word segment based on the first word frequency result, the second word frequency result and a synonym dictionary, wherein the synonym dictionary is used to find synonyms of the first word segment.

[0018] By adopting the above technical solution, by introducing a synonym dictionary to search for synonyms, if the word segmentation or synonyms of the source language text do not appear in the back-translation result, or the word frequency number of the word segmentation or synonym is different from the word frequency number of the word segmentation in the back-translation result, it is identified as a suspected abnormal translation word segmentation, which can reduce the number of false positives of abnormal translation word segmentation and improve the accuracy of abnormal translation word segmentation detection.

[0019] In a possible implementation, the first abnormal translation participle is screened based on the translation necessity evaluation rule, including: scoring the first abnormal translation participle based on the translation necessity evaluation rule; and determining the first abnormal translation participle whose scoring result is greater than a preset score as the second abnormal translation participle.

[0020] By adopting the above technical scheme, the candidate abnormal translation participles are scored based on the translation necessity evaluation rules. For example, for the part-of-speech rule, the importance of semantic carrying can be scored based on the part-of-speech of the abnormal translation participle. The higher the importance, the higher the score. For the scene association rule, the score can be based on whether it is associated with the user interface scene. The higher the association, the lower the score. For the target language style rule, the score can be based on whether it conforms to the target language style. If it does not conform to the target language style, a lower score can be set. By screening the candidate abnormal translation participles with scores greater than the preset score as the final output abnormal translation participle, it is possible to accurately screen out the missed / wrongly translated participles in the source language text of the user interface.

[0021] In a possible implementation, the first abnormal translation segmentation is screened based on the translation necessity evaluation rule, including: scoring and screening the first abnormal translation segmentation based on the part-of-speech rule to obtain a first screening result; scoring and screening the segmentation in the first screening result based on the scene association rule to obtain a second screening result; scoring and screening the segmentation in the second screening result based on the target language style rule to obtain a second abnormal translation segmentation.

[0022] By adopting the above technical solution, the candidate abnormal translation word segmentations are scored and screened by sequentially executing the part-of-speech rules, scene association rules and target language style rules, which can minimize the number of false positives of abnormal translation word segmentations and accurately screen out missed / wrongly translated word segmentations in the source language text of the user interface.

[0023] In a possible implementation, after obtaining the first screening result, it also includes: if the number of word segments in the first screening result is less than a preset quantity threshold, the word segmentation in the first screening result is used as the second abnormal translation word segmentation; the first screening result is screened based on the scene association rule, including: if the number of word segments in the first screening result is greater than or equal to the preset quantity threshold, the first screening result is screened based on the scene association rule.

[0024] By adopting the above technical scheme, it is also possible to select some or all of the rules in the translation necessity evaluation rules for execution according to a preset screening number threshold of abnormal translation segmentations. If the number of segmentations in the first screening result is less than the preset number threshold, it indicates that the number of abnormal translation segmentations obtained by screening is relatively small, that is, the number of false alarms is also relatively small. The electronic device can end the scoring and screening, and use the segmentations in the first screening result as the final output abnormal translation segmentations. If the number of segmentations in the first screening result is greater than or equal to the preset number threshold, it indicates that the number of abnormal translation segmentations obtained by screening is relatively large and further screening is required. The electronic device can continue to score and screen the segmentations in the first screening result based on the scene association rule to obtain a second screening result, which can reduce the number of abnormal translation segmentations finally output.

[0025] In a possible implementation, after obtaining the second screening result, it also includes: if the number of segmentations in the second screening result is less than a preset number threshold, the segmentations in the second screening result are used as second translation abnormal segmentations; the second screening result is screened based on the target language style rules, including: if the number of segmentations in the second screening result is greater than or equal to the preset number threshold, the second screening result is screened based on the target language style rules.

[0026] By adopting the above technical scheme, if the number of word segments in the second screening result is less than the preset number threshold, it indicates that the number of abnormal translation word segmentations obtained by the secondary screening is relatively small, that is, the number of false alarms is also relatively small. The electronic device can end the scoring and screening, and use the word segmentations in the second screening result as the final output abnormal translation word segmentations. If the number of word segmentations in the second screening result is greater than or equal to the preset number threshold, it indicates that the number of abnormal translation word segmentations obtained by screening is still relatively large and needs further screening. The electronic device can continue to score and screen the word segmentations in the second screening result based on the target language style rules to obtain the final abnormal translation word segmentations to be output, which can reduce the number of abnormal translation word segmentations finally output.

[0027] In a possible implementation, the part-of-speech rule includes: if the source language text includes participles of multiple parts of speech, scoring is performed based on the part-of-speech of the participles, wherein the scoring weight of a content word is greater than the scoring weight of a function word.

[0028] By adopting the above technical scheme, the words obtained by word segmentation processing are divided into content words and function words. If the user interface terminology includes both content words and function words, the content words have a relatively higher degree of semantic carrying in the text than the function words. The scoring weight of the content words is set to be greater than the scoring weight of the function words. The higher the semantic carrying degree of the words, the higher the score. The screening rule is to screen out candidate translation abnormal word segmentations with scores greater than the preset score as the final output translation abnormal word segmentations. That is, for user interface terms that include both content words and function words, if the translation abnormal word is a content word, it has a greater probability of being screened out than a function word.

[0029] In one possible implementation, the scene association rules include: scoring based on the correlation between the word segmentation and the theme of the user interface where the source language text is located, and based on the correlation between the word segmentation and the name of the application to which the user interface belongs, the correlation is negatively correlated with the scoring score, and the scoring weight of the theme of the user interface is greater than the scoring weight of the name of the application.

[0030] By adopting the above technical solution, by obtaining the correlation between the candidate translation abnormal participle and the theme of the user interface, as well as the correlation with the name of the application to which the user interface belongs, the higher the correlation, the greater the probability of being omitted, and the lower the corresponding score, and the screening rule is to screen out the candidate translation abnormal participle with a score greater than a preset score as the final output translation abnormal participle, that is, the omitted translation participle has a lower probability of being screened out, which can reduce the probability of the omitted translation participle being finally identified as the translation abnormal participle.

[0031] In a possible implementation, the target language style rule includes: scoring based on whether the word segmentation conforms to the target language style.

[0032] By adopting the above technical scheme, the candidate abnormal translation segmentation words are scored by judging whether they conform to the target language style. If they do not conform to the target language style, a lower score can be set, and the screening rule is to screen out the candidate abnormal translation segmentation words with a score greater than a preset score as the final output abnormal translation segmentation words, that is, the greater the probability that the segmentation words that do not conform to the target language style are omitted, the lower the corresponding score, and the lower the probability of being screened out, which can reduce the probability that the segmentation words that do not conform to the target language style are omitted and finally identified as abnormal translation segmentation words.

[0033] In a possible implementation, after outputting the screened second translation abnormal segmentation and the source language text corresponding to the second translation abnormal segmentation, the translation detection method further includes: obtaining the inspection result of the second translation abnormal segmentation, and adding the second translation abnormal segmentation whose inspection result is a false detection and the source language text corresponding to the second translation abnormal segmentation to a whitelist; and adding the whitelist to the translation necessity assessment rule.

[0034] By adopting the above technical scheme, the abnormal translation segmentation words and the source language text corresponding to the abnormal translation segmentation words that are finally output can be manually (translator) checked to see if there are indeed translation abnormalities. If there are false positive segmentations in the abnormal translation segmentation words that are finally output by manual inspection, these false positive segmentations and the corresponding source language texts can be added to a whitelist, so that subsequent electronic devices can improve the filtering ability of false positive segmentations based on the updated whitelist. The whitelist can also be integrated into the translation necessity evaluation rules, or it can be independent of the translation necessity evaluation rules. By setting a false positive whitelist, the accuracy of translation abnormal segmentation detection can be further improved, and the missed / wrongly translated segmentations in the source language text of the user interface can be accurately screened out.

[0035] In a second aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory is used to store program instructions; the processor is used to read and execute the program instructions stored in the memory, and when the program instructions are executed by the processor, the electronic device executes the translation detection method of the first aspect, or executes the translation detection method of the second aspect.

[0036] In a third aspect, the present application provides a computer-readable storage medium storing program instructions, which, when executed on an electronic device, causes the electronic device to execute the translation detection method of the first aspect, or execute the translation detection method of the second aspect.

[0037] In a fourth aspect, the present application provides a chip coupled to a memory in an electronic device, the chip being used to control the electronic device to execute the translation detection method of the first aspect, or to execute the translation detection method of the second aspect.

[0038] In addition, the technical effects brought about by the second to fourth aspects can be found in the descriptions related to the methods of each design in the above method part, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1A and 1B This is a schematic diagram of the interface before and after the language is switched in the interface language of the user equipment provided by an embodiment of the present application.

[0040] Figure 2A and 2B This is a schematic diagram of the interface before and after the language is switched in the interface terms of a user device provided by another embodiment of the present application.

[0041] Figure 3A and 3B This is a schematic diagram of the interface before and after the language is switched on the dial interface of a user device provided by an embodiment of the present application.

[0042] Figure 4 It is a scenario diagram of a translation detection system provided in an embodiment of the present application.

[0043] Figure 5 It is a flowchart of a translation detection method provided in an embodiment of the present application.

[0044] Figure 6 It is a schematic diagram of the rule content of the translation necessity evaluation rule provided in one embodiment of the present application.

[0045] Figure 7 It is a schematic diagram of omitted translation of source language text in a user interface provided by an embodiment of the present application.

[0046] Figure 8 This is a software architecture diagram of a user interface terminology detection tool provided in one embodiment of the present application.

[0047] Fig. 9 It is a flowchart of a translation detection method provided in another embodiment of the present application.

[0048] Fig.10 It is a flowchart of a translation detection method provided in yet another embodiment of the present application.

[0049] Fig.11 This is a hardware architecture diagram of an electronic device provided in one embodiment of the present application.

[0050] Fig.12 This is a hardware architecture diagram of a user device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0051] To facilitate understanding, some illustrations of concepts related to the embodiments of the present application are given by way of example for reference.

[0052] It should be noted that the terms "first" and "second" involved in the embodiments of the present application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field in this application. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. It should be understood that, unless otherwise specified in this application, " / " means or. For example, A / B can represent A or B. "And / or" in this application is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. "At least one" means one or more. "Multiple" means two or more than two. For example, at least one of a, b or c can represent: a, b, c, a and b, a and c, b and c, a, b and c seven situations.

[0054] In order to meet the needs of multiple languages, user devices usually set texts in multiple languages ​​for the same user interface term before leaving the factory, so that texts in different languages ​​can be displayed according to the needs of different users. During the development process of user devices, it is necessary to detect the multilingual texts corresponding to each user interface term to avoid missing or wrong translation problems that affect the user experience.

[0055] The user device in the embodiment of the present application may be a mobile phone, a portable android device (Pad), a personal digital assistant (PDA), a wearable device (such as a smart watch), a monitoring device, a vehicle-mounted device, or other electronic device. Windows Or devices with other operating systems, which is not limited in this application.

[0056] The user interface (UI) in the embodiments of the present application is a medium interface for interaction and information exchange between an application or operating system and a user, and can realize the conversion between the internal form of information and the form acceptable to the user (such as a visual operating environment). The term user interface may refer to the text contained in the user interface, and the text language may be Chinese, English, Japanese, Korean, Arabic, French, Russian, German, Portuguese, etc. The user interface of an application may be a source code written in a specific computer language such as JAVA or extensible markup language (XML). The interface source code is parsed and rendered on the user device, and finally presented as content that the user can recognize, such as pictures, text, buttons and other controls. Controls are the basic elements of the user interface. Typical controls include buttons, widgets, toolbars, menu bars, text boxes, scroll bars, images and text. The properties and contents of controls in the interface are defined by tags or nodes, such as XML through <textview> 、 <imgview> 、 <videoview>The controls contained in the interface are specified by nodes such as . A node corresponds to a control or attribute in the interface. After parsing and rendering, the node is presented as user-visible content. In addition, many applications, such as hybrid applications, usually also contain web pages in their interfaces. A web page, also known as a page, can be understood as a special control embedded in the application interface. A web page is source code written in a specific computer language, such as hypertext markup language (HTML), cascading style sheets (CSS), JAVA scripts (JavaScript, JS), etc. The web page source code can be loaded and displayed as user-recognizable content by a browser or a web page display component with similar functions to a browser. The specific content contained in a web page is also defined by tags or nodes in the web page source code. For example, HTML is defined by 、 、 <video> 、 <canvas>To define the elements and attributes of a web page. The display interface, application program interface, game interface and other interfaces in the embodiments of the present application can all be referred to as user interfaces.

[0057] The most common form of user interface is graphical user interface (GUI), which refers to a user interface related to computer operation that is displayed in a graphical manner. It can be an icon, window, control or other interface element displayed on the display screen of the user device.

[0058] In some embodiments, the user device may install one or more APPs (Applications). APPs may be referred to as applications, which are software programs that can implement one or more specific functions. For example, instant messaging applications, video applications, audio applications, image capture applications, cloud desktop applications, sports and health applications, etc. The applications mentioned in the following embodiments may be system applications installed when the user device leaves the factory, or may be third-party applications downloaded by the user from the Internet or obtained from other user devices during the use of the user device.

[0059] like Figure 1A As shown, taking the user device 200 as a mobile phone as an example, the user interface currently displayed by the user device 200 is a setting interface 201, and the language of the interface terms is simplified Chinese. The setting interface 201 displays multiple setting options, such as a "biometric identification and password" setting option, an "application" setting option, a "battery" setting option, a "storage" setting option, a "security" setting option, a "privacy" setting option, etc.

[0060] If the language of the interface is switched from Simplified Chinese to English, the content displayed in the setting interface 201 changes to Figure 1B Settings is changed to "Settings", "Biometrics & Password" setting option is changed to "Biometrics & Password" setting option, "Application" setting option is changed to "Apps" setting option, "Battery" setting option is changed to "Battery" setting option, "Storage" setting option is changed to "Storage" setting option, "Security" setting option is changed to "Security" setting option, and "Privacy" setting option is changed to "Privacy" setting option.

[0061] like Figure 2A As shown, taking the user device 200 as a mobile phone as an example, the user interface currently displayed by the user device 200 is a time selection control 202, and the language of the interface language is English. If the language of the interface language is switched from English to Portuguese, the content displayed in the time selection control 202 changes to Figure 2B shown.

[0062] exist Figure 2B In the example, the Portuguese content displayed in the time selection control 202 does not include Figure 2A The Portuguese translation of the English content "From" in the time selection control 202 is "Theend", which is quite different from the English content "From". Incorrect translation will affect the user's correct use of related functions, and further affect the user experience of the device or APP.

[0063] like Figure 3A As shown, taking the user device 200 as a smart watch as an example, the user interface currently displayed by the user device 200 is a dial interface 203, and the language of the interface terms is English. The contents displayed on the dial interface 203 include: HEART RATE 97bpm, CALORIES 328, ALTITUDE 2680m, WEATHER 24°, VO2 MAX 58, HOURS ACTIVE 9hours, ACTIVE 16mins, STEPS 7645, and FRI 18.

[0064] If the language of the interface is switched from English to Simplified Chinese, the content displayed in the dial interface 203 changes to Figure 3B As shown, the contents displayed on the dial interface 203 include: heart rate 97bpm, calories 328, altitude 2680m, weather 24°, maximum oxygen uptake 58, active time 9 hours, exercise time 16 minutes, steps 7645 and Friday the 18th.

[0065] Since the user device 200 supports a wide variety of languages, the content displayed in the user interface may need to be translated into texts corresponding to nearly a hundred languages. For the user device 200 including a large number of user interfaces, the workload of detecting the translation accuracy of the user interface terms is large and the detection difficulty is high.

[0066] Based on this, in order to accurately detect errors in the translation of terms in the user interface (such as missing translations / wrong translations) and improve the accuracy of multilingual terms in the user interface, the present application provides a translation detection method. The method can be implemented in the electronic device provided in the present application, and the electronic device can be a development end device (device used by developers or testers), such as a computer, a server, etc., to detect the user interface terms before the user device leaves the factory, or after the user device leaves the factory, update the user interface terms by software updates to correct the translation errors in the user interface terms.

[0067] See also Figure 4 , is a scenario diagram of a translation detection system provided in an embodiment of the present application, and the translation detection system may include a verification server 11 and a data server 12 that provides various data supports. The data server 12 is used to provide language pairs (source language-target language, such as Chinese-English, English-French, etc.), text to be translated (source language text), translation corresponding to the text to be translated (target language text), user interface where the source language text is located, application description information of the user interface, etc.

[0068] The verification server 11 is used to detect the translation corresponding to the to-be-translated text based on the content provided by the data server 12, and determine whether the translation has translation errors (omissions or mistranslations). The verification server 11 can output a list of translations with translation errors, and the list of translations can include one or more translations with translation errors. The verification server 11 can also output a word set with omissions / mistranslations in each translation, and the word set can include one or more words with omissions / mistranslations.

[0069] It should be noted that Figure 4 The system scenario diagram shown is only an example. The server and scenario described in the embodiment of the present application are intended to more clearly illustrate the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. Ordinary technicians in this field can know that with the evolution of the system and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0070] See also Figure 5 , is a flow chart of the translation detection method provided in the embodiment of the present application. The translation detection method can be applied to electronic devices, and the electronic devices can be computers, servers and other devices. The translation detection method can include the following steps:

[0071] S501: Acquire data to be detected, where the data to be detected includes a source language text in a user interface, a translation corresponding to the source language text, and user interface information of the source language text.

[0072] The number of data to be detected can be determined according to actual needs, and the embodiments of the present application do not limit this. For example, the source language text in the data to be detected can refer to the source language text that appears in all user interfaces in the user device and needs to be translated, and the translation corresponding to the source language text can be the translation of all languages ​​supported by the user device, and the translation can be obtained by machine translation model translation or manual translation. For example, the source language text is English, and the content is "health". If the user device also supports Simplified Chinese, Traditional Chinese, Japanese, Korean, Arabic, French, Russian, German, Portuguese and other N languages, the translation corresponding to the English "health" includes the translations corresponding to the N languages ​​respectively.

[0073] Among the user interface terms of existing user devices, the user interface terms in Chinese and English are relatively mature, and there are basically no problems of missed or wrong translation. That is, the language of the source language text can be set to Chinese or English, and the electronic device detects whether there are any translation anomalies in the translation of the source language text into other languages.

[0074] In some embodiments, since English has similarities with many other languages, setting the source language text to English can further improve the efficiency and accuracy of translation detection and reduce the workload of translation detection.

[0075] The user interface information of the source language text may include information of the user interface where the source language text is located and information of the application to which the user interface belongs. For example, the user interface information may include information such as the usage scenario and theme of the user interface, and the application information to which the user interface belongs may include information such as the type and name of the application.

[0076] In some embodiments, the data to be detected can be pre-stored in a designated storage area, which can be the memory of the electronic device itself or the memory of other electronic devices (such as a server or database that can communicate with the electronic device). The electronic device can obtain the data to be detected by reading the data to be detected from the designated storage area or receiving the data to be detected sent by other electronic devices.

[0077] For example, if the source language text is English, the source language text may include tens of thousands of pieces, and the number of translations corresponding to each source language text is related to the languages ​​supported by the user device. Assuming that the user device supports 80 languages ​​(including English), the number of translations corresponding to each source language text is 79 (needing to be translated into 79 languages ​​other than English).

[0078] S502: Back-translate the translation to obtain a first back-translation result corresponding to the translation.

[0079] In some embodiments, the language of the first back-translation result is the same as the language of the source language text, and the electronic device can call a preset translation model to back-translate the translation to obtain a first back-translation result corresponding to the translation. The translation model can be pre-installed in the electronic device, and the electronic device can also call the translation model located on the cloud server by accessing the network. The translation model can be an existing machine translation model (such as a machine translation model provided by open source or a third party), or it can be a machine translation model developed by the user device manufacturer, which is not limited in the embodiments of the present application. In order to improve the accuracy and efficiency of translation detection, the translation model called by the electronic device can select a machine translation model with relatively high translation accuracy.

[0080] S503: Perform word segmentation processing on the translation to obtain one or more word segments, and construct a back-translation word set corresponding to the word segments.

[0081] In some embodiments, the electronic device may use an existing word segmentation algorithm to perform word segmentation processing on the translation to obtain one or more word segments, and the embodiment of the present application does not limit the word segmentation algorithm. For example, the word segmentation algorithm may include a forward (reverse) maximum matching method, a best matching method, a word-by-word traversal method, or a word frequency statistics method.

[0082] For example, "Manage screen time and protect security remotely" is parsed into multiple participles such as "Manage", "screen", "time", "and", "protect", "security", and "remotely".

[0083] The back-translation word set may include the back-translation word corresponding to each segmented word (the language of the back-translation word is the same as the language of the source language text), and the number of back-translation words corresponding to each segmented word may include one or more. For example, a bilingual dictionary and / or a bilingual term list may be used to search for the back-translation word corresponding to each segmented word, and the back-translation word set summarizes the search results of the back-translation word corresponding to each segmented word. The bilingual term list may be an existing public bilingual term comparison table, or a bilingual term comparison table accumulated by the user device manufacturer itself.

[0084] For example, a bilingual dictionary and a bilingual term list are used to search for back-translated words corresponding to each segmentation, the translation is segmented to obtain segmentation A1, the bilingual dictionary is searched to obtain back-translated words B1 corresponding to segmentation A1, and the bilingual term list is searched to obtain back-translated words B1 and B2 corresponding to segmentation A1. The search results of the bilingual dictionary and the bilingual term list (back-translated words B1 and B2) can be summarized and added to the back-translated word set.

[0085] Taking the source language text as English and the translation as Portuguese as an example, the language of the word segmentation is Portuguese, the language of the back-translated words is English, the bilingual dictionary is a Portuguese-English dictionary, and the bilingual terminology table is a Portuguese-English terminology comparison table. The back-translated words corresponding to each word segmentation can be obtained by searching the Portuguese-English dictionary and the Portuguese-English terminology comparison table, and then a back-translated word set is constructed based on the found back-translated words.

[0086] In some embodiments, step S502 and step S503 may be executed in parallel, or step S502 may be executed first and then step S503, or step S503 may be executed first and then step S502.

[0087] S504: constructing a second back-translation result based on the first back-translation result and the back-translation word set, and calculating the similarity between the second back-translation result and the corresponding source language text.

[0088] In some embodiments, each source language text corresponds to a translation, each translation corresponds to a first back-translation result, the first back-translation result is associated with the second back-translation result, and the source language text corresponding to the second back-translation result is the source language text corresponding to the first back-translation result.

[0089] The electronic device may use an existing text similarity calculation algorithm to calculate the similarity between the second back-translation result and the source language text. The embodiment of the present application does not limit the text similarity calculation algorithm. For example, the text similarity calculation algorithm may include a cosine similarity calculation algorithm, a similarity calculation algorithm based on Euclidean distance, a similarity calculation algorithm based on a latent semantic indexing (LSI) model, a text similarity calculation algorithm based on a bag-of-words model, a Jaccard similarity calculation algorithm, and the like.

[0090] In some embodiments, the similarity between the second back-translation result and the corresponding source language text may also refer to semantic similarity, and the electronic device may further calculate the semantic similarity between the second back-translation result and the corresponding source language text based on a semantic similarity calculation algorithm. The embodiment of the present application does not limit the semantic similarity calculation algorithm, for example, the semantic similarity calculation algorithm may include a Word2Vec algorithm, a SimHash algorithm, etc.

[0091] In some embodiments, the electronic device may include one or more back-translated texts based on the second back-translated result obtained by constructing the first back-translated result and the back-translated word set. For example, the electronic device may replace the word in the first back-translated result (the word has the same or similar semantics as the back-translated word) with the back-translated word in the back-translated word set to obtain one or more back-translated texts. Since the user interface terms are mostly short sentences (most of the user interface terms contain only one or two words, and a few are three or more words), and the literary quality is relatively low, most words can obtain more accurate translations through dictionaries and glossaries. The embodiment of the present application back-translates the word segmentation of the translation by adding a bilingual dictionary and a bilingual glossary to obtain a back-translated word set, and then constructs the second back-translated result based on the first back-translated result and the back-translated word set, which can supplement the first back-translated result and reduce the back-translation accuracy loss that may be caused by back-translating the translation using a machine translation model.

[0092] In some embodiments, the electronic device may calculate the similarity between each back-translated text and the source language text, and use the maximum similarity among the calculated multiple similarities as the similarity between the second back-translated result and the source language text. In other embodiments, the electronic device may also use the average or median value of the multiple similarities as the similarity between the second back-translated result and the source language text.

[0093] For example, the source language text is: "sound&vibration", the first back-translation result is: "sound&vibration", the back-translation word set is: [(sound,voice),(vibration)], and the second back-translation results constructed by the electronic device based on the first back-translation result and the back-translation word set include: "sound&vibration" and "voice&vibration".

[0094] In some embodiments, since the first back-translation result and the back-translation word set are both associated with the translation, the similarity between the second back-translation result and the corresponding source language text can be used to characterize the translation accuracy of the translation. The higher the similarity between the second back-translation result and the corresponding source language text, the higher the accuracy of the translation obtained based on the source language text. Conversely, the lower the similarity between the second back-translation result and the corresponding source language text, the lower the accuracy of the translation obtained based on the source language text.

[0095] S505: Based on the similarity between the second back-translation result and the corresponding source language text, a text set suspected of having translation anomalies is obtained by screening.

[0096] In some implementations, a similarity threshold may be pre-set, and the calculated similarity may be compared with the similarity threshold to screen out the second back-translation results and source language texts with suspected translation anomalies. The electronic device may construct a text set with suspected translation anomalies based on the screened out second back-translation results and source language texts with suspected translation anomalies. For example, the electronic device may filter out the second back-translation results and the corresponding source language texts with similarities greater than or equal to the similarity threshold, and the text set includes text pairs of the second back-translation results and the corresponding source language texts with similarities less than the similarity threshold.

[0097] In some embodiments, the similarity threshold can be set according to actual detection requirements, and the embodiments of the present application do not limit this. For example, the similarity value range is 0 to 1 (the closer the value is to 1, the higher the similarity), and the similarity threshold can be set to 0.9.

[0098] If the similarity between a second back-translation result and the corresponding source language text is greater than or equal to the similarity threshold, it indicates that the accuracy of the translation obtained based on the source language text translation meets the requirements, and it can be considered that there are no translation errors (omissions or mistranslations) in the translation, and there is no need to process the second back-translation result of the translation and the source language text corresponding to the translation. If the similarity between a second back-translation result and the corresponding source language text is less than the similarity threshold, it indicates that the accuracy of the translation obtained based on the source language text translation does not meet the requirements, and it can be considered that there are suspected translation errors (omissions or mistranslations) in the translation, and the second back-translation result and the corresponding source language text can be added to a preset text set, for example, a text set suspected of translation anomalies, so as to facilitate the subsequent steps to continue the analysis and processing.

[0099] Since user interface terms are mostly short sentences with low literary value, the translation accuracy of the translations obtained based on machine translation models or human translators is relatively high, and there are only a few user interface terms with translation anomalies. By filtering the second back-translation results with a similarity greater than or equal to a similarity threshold and the corresponding source language text, the electronic device can reduce the amount of data processing, save computing resources, and shorten the detection path of these texts without anomalies.

[0100] In some embodiments, the electronic device may further perform a similarity score based on the calculated similarity, and add text pairs (the second back-translation result and the corresponding source language text) with scores below a threshold to a set of texts suspected of having translation anomalies.

[0101] S506: Perform word segmentation processing and word frequency statistics processing on the second back-translation result in the text set and the corresponding source language text, respectively, to obtain a first word frequency result of the second back-translation result and a second word frequency result of the source language text.

[0102] In some embodiments, for each second back-translation result in the text set and the source language text corresponding to each second back-translation result, the electronic device may use an existing word segmentation algorithm to perform word segmentation processing on the second back-translation result to obtain one or more first word segments, and perform word segmentation processing on the source language text to obtain one or more second word segments. The embodiment of the present application does not limit the word segmentation algorithm. For example, the word segmentation algorithm may include a forward (reverse) maximum matching method, a best matching method, a word-by-word traversal method, or a word frequency statistics method.

[0103] After completing the word segmentation process, the electronic device can also use an existing word frequency statistics algorithm to obtain the word frequency of each word segmentation. The embodiment of the present application does not limit the word frequency statistics algorithm. For example, the word frequency statistics algorithm can be a term frequency-inverse document frequency (TF-IDF) algorithm.

[0104] Taking the TF-IDF algorithm as an example, the electronic device can use the TF-IDF algorithm to perform word frequency statistics on the first word segmented by word segmentation to obtain a first word frequency result, and use the TF-IDF algorithm to perform word frequency statistics on the second word segmented by word segmentation to obtain a second word frequency result.

[0105] For example, the second back-translation result is segmented, and the first segmented words obtained include "health" and "exercise". "health" appears once in the second back-translation result, and "exercise" appears twice in the second back-translation result. The first word frequency result is: [health 1, exercise 2]. The source language text is segmented, and the second segmented words obtained include "health" and "activity". "Health" appears once in the source language text, and "activity" appears twice in the source language text. The second word frequency result is: [health 1, activity 2].

[0106] In some embodiments, if the second back-translation result includes multiple back-translated texts, the electronic device can use an existing word segmentation algorithm to perform word segmentation processing on each back-translated text separately, and obtain word frequency results of each back-translated text separately, that is, the first word frequency result can include multiple sub-word frequency results, and each sub-word frequency result is the result obtained by performing word segmentation processing and word frequency statistics processing on a back-translated text.

[0107] In some embodiments, after the electronic device performs word segmentation processing on the second back-translation result or the source language text, it can also perform part-of-speech tagging on each word segmentation to obtain the part-of-speech information of each word segmentation, so as to facilitate the subsequent evaluation of translation necessity based on the part-of-speech information of the candidate abnormal translation word segmentation. For example, the part-of-speech information may include verbs, nouns, adjectives, adverbs, articles, conjunctions, prepositions, pronouns, etc.

[0108] S507: Determine a first translation abnormal word segment in the source language text based on the first word frequency result and the second word frequency result.

[0109] After obtaining the first word frequency result and the second word frequency result, the first word frequency result and the second word frequency result can be compared to determine the first abnormal translation segmentation in the source language text. The first abnormal translation segmentation is a candidate abnormal translation segmentation, and some of the first abnormal translation segmentations may be segmentations that are indeed abnormal translations, and some segmentations that have been correctly translated. They can be subsequently filtered based on the translation necessity evaluation rules to reduce the number of false positives.

[0110] For example, the second back-translation result includes a back-translation text, that is, the first word frequency result is a group of word frequency results, and the first word frequency result is compared with the second word frequency result. For each word segment in the source language text, if it does not appear in the first word frequency result, or the word frequency number is different from the word frequency number recorded in the first word frequency result, the word segment is used as the first translation abnormal word segment.

[0111] For example, the second frequency result is: [health 1, activity 1], and the first frequency result is: [health 1]. Since "activity" in the source language text does not appear in the first frequency result, "activity" in the source language text can be used as the first abnormal translation segmentation. For example, the second frequency result is: [health 1, activity 1], and the first frequency result is: [health 1, activity 2]. Since "activity" appears once in the source language text and twice in the second back translation result, the two are different, "activity" in the source language text is used as the first abnormal translation segmentation.

[0112] In some embodiments, in order to reduce the number of false positives of abnormal translation word segmentation, a synonym dictionary may be introduced for retrieval, that is, the electronic device may determine the first abnormal translation word segmentation in the source language text based on the first word frequency result, the second word frequency result and the synonym dictionary.

[0113] For example, if a synonym dictionary is not introduced, the second frequency result is: [health 1, activity 1], and the first frequency result is [health 1, exercise1]. Since "activity" in the source language text does not appear in the first frequency result, the electronic device will use "activity" in the source language text as the first abnormal translation segmentation. If a synonym dictionary is introduced, although "activity" in the source language text does not appear in the first frequency result, the electronic device can determine that "activity" and "exercise" are synonyms by querying the synonym dictionary, and "exercise" appears once in the second back translation result, and the two have the same frequency. The electronic device will not use "activity" in the source language text as the first abnormal translation segmentation.

[0114] If the second back-translation result includes multiple back-translated texts, that is, the first word frequency result includes multiple subword frequency results, each subword frequency result can be compared with the second word frequency result respectively. For each word segment in the source language text, if it does not appear in all subword frequency results, or the word frequency number is different from the word frequency number recorded in all subword frequency results, the word segment shall be regarded as the first abnormal translation word segment.

[0115] For example, the second back-translation result includes two back-translated texts (hereinafter referred to as the first back-translated text and the second back-translated text), the word frequency result of the first back-translated text (hereinafter referred to as the first sub-word frequency result) is: [exercise 1], the word frequency result of the second back-translated text (hereinafter referred to as the second sub-word frequency result) is: [activity 1], and the second word frequency result is: [health 1,activity 1]. The word frequency number of "activity" in the source language text is the same as the word frequency number of "activity" in the second back-translated text, and "health" does not appear in the first sub-word frequency result or in the second sub-word frequency result, that is, the electronic device will take "health" in the source language text as the first translation abnormal word segmentation.

[0116] If the first subword frequency result is: [health 2, exercise 1], the second subword frequency result is: [health 2, activity 1], and the second word frequency result is: [health 1, activity 1], the word frequency number of "activity" in the source language text is the same as the word frequency number of "activity" in the second back-translated text, while the word frequency number of "health" is different from the word frequency number of "health" in the first or second back-translated text, that is, the electronic device can use "health" in the source language text as the first abnormal translation word.

[0117] S508: Scoring the first abnormal translation segmentations based on the translation necessity evaluation rule, and screening out the second abnormal translation segmentations according to the scoring results of each first abnormal translation segmentation.

[0118] In some embodiments, the translation necessity assessment rule may include at least one of a part-of-speech rule, a scene association rule, and a target language style rule. The translation necessity assessment rule may be pre-stored in an electronic device or a cloud server, and the electronic device may score each first translation abnormal participle based on the translation necessity assessment rule, and then may screen the first translation abnormal participle with a score higher than a preset score according to the scoring result of each first translation abnormal participle, and these first translation abnormal participles with scores higher than the preset score are identified as participles (second translation abnormal participles) that are indeed abnormal in translation. The preset score can be set according to actual needs, and the embodiment of the present application is not limited to this.

[0119] The idea of ​​constructing translation necessity assessment rules is to construct a rule set based on the characteristics of linguistics, translation studies and user interface terminology translation. The rule set can be stored in XML format. For example, the rule set includes part-of-speech rules, scene association rules and target language style rules. The execution order can be part-of-speech rules->scene association rules->target language style rules. Part-of-speech rules operate faster and have a relatively high recognition accuracy, so they are executed first. Scene association rules involve image processing, which are relatively slow to operate, and have a higher recognition accuracy than target language style rules, so they are executed second. The recognition accuracy of target language style rules is the lowest, so they are executed third. In actual applications, the execution order of each screening rule can also be adjusted according to needs.

[0120] The electronic device can also select to execute some or all of the rules in the rule set according to the preset threshold value of the number of the second translation abnormal participles. For example, if the preset threshold value is 200, if the electronic device scores and screens each first translation abnormal participle through the part-of-speech rule, if the number of the second translation abnormal participles screened is less than 200, the electronic device ends the scoring and screening, and outputs the screened second translation abnormal participles; if the number of the second translation abnormal participles screened is greater than 200, the electronic device continues to execute the scene association rule to continue to score and screen the second translation abnormal participles screened by the part-of-speech rule, if the number of the second translation abnormal participles screened again is less than 200, the electronic device ends the scoring and screening, and outputs the second translation abnormal participles screened by the scene association rule; if the number of the second translation abnormal participles screened again is still greater than 200, the electronic device continues to execute the target language style rule to continue to score and screen the second translation abnormal participles screened by the scene association rule, and outputs the second translation abnormal participles screened by the target language style rule. That is, if the number of abnormal second translation segmentations obtained through screening by the target language style rules is greater than 200, since all screening rules have been executed, the electronic device will still end the scoring and screening, and output the abnormal second translation segmentations obtained through screening by the target language style rules. The number of abnormal second translation segmentations finally outputted can be reduced by adding a whitelist later.

[0121] User interface terms are mostly short sentences with relatively simple sentence structures. Different scoring weights can be assigned to different parts of speech according to the importance of the parts of speech of each word in the source language text to the semantics. Figure 6 As shown, the part-of-speech rules of an embodiment of the present application are illustrated. Words can be divided into two major categories: content words and function words. In the part-of-speech rules, if the source language text includes both content words and function words, the scoring weight of the content words can be set to be greater than the scoring weight of the function words. For example, content words can include multiple parts of speech such as nouns, verbs, adjectives, numerals, quantifiers, pronouns, etc., and function words can include multiple parts of speech such as articles, conjunctions, prepositions, etc.

[0122] If the source language text includes content words of multiple parts of speech, or function words of multiple parts of speech. For various parts of speech in content words or function words, the setting of scoring weights can also be continued. For example, in content words, the scoring weight of nouns is greater than the scoring weight of verbs, and the scoring weight of verbs is greater than the scoring weight of adjectives. In function words, the scoring weight of articles is greater than the scoring weight of conjunctions, and the scoring weight of conjunctions is greater than the scoring weight of prepositions. It can be understood that the scoring weights of various parts of speech in content words can be set according to actual needs, and the scoring weights of various parts of speech in function words can also be set according to actual needs, and the embodiments of the present application are not limited to this.

[0123] For example, the baseline score for the first abnormal translation segmentation based on the part-of-speech rule is 10 points. Assuming that the source language text tx1 includes noun w1 and conjunction w2, the score weight of noun w1 is 0.6, and the score weight of conjunction w2 is 0.4, based on step S507, it is determined that the first abnormal translation segmentation in the source language text tx1 is conjunction w2, and the electronic device scores conjunction w2 based on the part-of-speech rule as 10*0.4=4 points.

[0124] Assume that the source language text tx2 includes noun w3. Since the source language text tx2 only includes participles of one part of speech, the scoring weight of noun w3 is 1. Based on step S507, it is determined that the first abnormal translated participle in the source language text tx2 is noun w3. The electronic device scores noun w3 based on the part of speech rule as 10*1=10 points.

[0125] Assume that the source language text tx3 includes preposition w4. Since the source language text tx3 only includes participles of one part of speech, the scoring weight of preposition w4 is 1. Based on step S507, it is determined that the first abnormal translation participle in the source language text tx3 is preposition w4. The electronic device scores preposition w4 based on the part of speech rule as 10*1=10 points.

[0126] Assume that the source language text tx4 includes noun w5, verb w6 and preposition w7, the scoring weight of noun w5 is 0.5, the scoring weight of verb w6 is 0.3, and the scoring weight of preposition w7 is 0.2. Based on step S507, it is determined that the first translation abnormal participle in the source language text tx4 is verb w6 and preposition w7. The electronic device scores verb w6 as 10*0.3=3 points and preposition w7 as 10*0.2=2 points based on the part-of-speech rule.

[0127] In some embodiments, when scoring the first translated abnormal word based on the part-of-speech rule, the word frequency of the first translated abnormal word may also be considered. The greater the difference between the word frequency of the first translated abnormal word and the word frequency of the back-translated text, the greater the scoring weight of the first translated abnormal word.

[0128] In some embodiments, the scene association rules can be formulated based on the user interface information of the source language text, which can effectively reduce the false positives of omitted translation of interface terms. Figure 6 As shown, the scene association rule of an embodiment of the present application is illustrated. The weight of the subject word of the user interface of the source language text is greater than the weight of the application name.

[0129] When preparing the data to be detected of the user device, the application name and the keywords of the user interface can be pre-added in the user interface information of the source language text. The application name is generally unique, and the application information installed on the user device (including the application icon, name, version number, etc.) can be obtained through the PackageManager class provided in the operating system of the user device. For the keywords of the user interface, the user interface can be identified by using the existing image semantic recognition model. The keywords of the user interface identified based on the image semantic recognition model can also be reviewed and adjusted by the developer to obtain the final keywords of the user interface.

[0130] For example, in the process of translating the source language text into the translation, the source language text can be compared with the subject words of the user interface, and the words in the source language text that are the same or similar to the subject words (such as synonyms or hyponyms) can be omitted in the translation process. That is, if the first translation abnormal participle is the same as the subject word or application name of the user interface, or is a synonym, or a hyponym, the first translation abnormal participle is more likely to be omitted, and can be given a lower score to facilitate filtering.

[0131] For the first translation abnormal participle, the first translation abnormal participle can also be calculated for the relevance with the subject word and application name of the user interface, for example, the point mutual information algorithm (PMI) can be used to calculate the relevance between the first translation abnormal participle and the subject word of the user interface, and the relevance between the first translation abnormal participle and the application name. If the first translation abnormal participle has a higher relevance with the subject word / application name of the user interface (indicating that the probability of being omitted is greater), the score of the first translation abnormal participle is lower. For example, the weight of the subject word of the user interface is 0.6, the weight of the application name is 0.4, and the score of the first translation abnormal participle w11 is 0.6*score1+0.4*score2, where score1 is a score based on the relevance between the first translation abnormal participle w11 and the subject word of the user interface, and score2 is a score based on the relevance between the first translation abnormal participle w11 and the application name.

[0132] In some embodiments, when scoring the scene association rules, the electronic device may also query the vocabulary network (Wordnet) to obtain synonyms, upper and lower concept words of the first abnormal translation participle, and then calculate the correlation with the subject words and application names of the user interface respectively. The higher the correlation, the lower the score. For example, the lowest score among the first abnormal translation participle, synonyms, upper concept words, and lower concept words can be used as the final score of the first abnormal translation participle.

[0133] like Figure 7 As shown, taking the case where the source language text is Chinese and the translation language is English as an example, user interface I1 is the user interface where the source language text is located, and user interface I2 is the user interface where the translation is located. The translation corresponding to the source language text "exercise time" is "Activity", that is, the participle "exercise" in the source language text is translated into "Activity", and the participle "duration" is omitted. User interface I1 displays the user's exercise time through a bar chart, that is, the subject word of user interface I1 is duration. In the process of translating from Chinese to English, the "duration" in the source language text "exercise time" can be omitted based on the subject word of user interface I1, that is, "exercise time" can be translated into "Activity", which is not a translation error.

[0134] In some embodiments, the target language style rules can be summarized from books related to the linguistics (vocabulary, grammar rules), translation style, etc. of the target language, and the target language style rules can be used to further reduce the false positives of omitted translation of interface terms. For example, the target language style rules can be constructed based on three aspects: vocabulary, grammar, and N-gram, which can delete words that do not conform to the target language style (thinking habits, language habits, and expressions) to avoid redundant translations, that is, these words that do not conform to the target language style can be omitted and do not constitute translation errors.

[0135] For example, for words that do not conform to the target language style, a lower score is given to facilitate filtering. For example, the target language style rule is a target language style set, which stores multiple style rules related to the target language, and each style rule can be related to a word. When scoring the target language style rule, the score can be based on whether the first translation abnormal word segmentation has appeared in the target language style set. If it has appeared, a relatively low score can be given, and if it has not appeared, a relatively high score can be given.

[0136] Taking the source text in Chinese and the translation in English as an example, the source text is "a bumper harvest in agriculture". If the word-by-word translation is performed, the corresponding translation is "good harvests in agriculture". However, "harvest" as a word for harvest can be determined to refer to the agricultural field, which does not conform to the language style of English. The translation corresponding to "a bumper harvest in agriculture" should be "good harvests", that is, the word "agriculture" in the source text is omitted, which is not a translation error.

[0137] S509: Outputting the second abnormal translation segmentation and the source language text corresponding to the second abnormal translation segmentation.

[0138] After the second abnormal translation participle is screened, the second abnormal translation participle and the source language text corresponding to the second abnormal translation participle can be output. The second abnormal translation participle and the source language text corresponding to the second abnormal translation participle can be output in the form of a table. For example, one column of the table is used to display the second abnormal translation participle, and the other column is used to display the source language text corresponding to the second abnormal translation participle. If the second abnormal translation participle is the source language text, only the second abnormal translation participle can be displayed.

[0139] In some embodiments, the second translation abnormal participle and the source language text corresponding to the second translation abnormal participle can be sent to a preset terminal, and the user of the preset terminal performs manual (e.g., translator) reconfirmation to determine whether there is a translation error. If the manual confirmation is a false alarm, a white list can be constructed based on the content of these false alarms, so that subsequent electronic devices can filter out the false alarm second translation abnormal participle based on the white list. The content in the white list can be updated according to the manual confirmation result, that is, the electronic device can screen and obtain the second translation abnormal participle based on the translation necessity assessment rule and the regularly updated white list.

[0140] See also Figure 8 , which is a software architecture diagram of the user interface terminology detection tool of the embodiment of the present application. The electronic device may be pre-installed with a user interface terminology detection tool 101, which detects multilingual texts corresponding to user interface terms in the user device to avoid missing or mistranslating problems that affect the user experience.

[0141] like Figure 8 As shown, the user interface terminology detection tool 101 can be divided into four levels: a data input level, a similarity calculation and screening level, a translation anomaly identification level, and a translation necessity assessment level.

[0142] The data input layer may include content acquisition, machine translation model and whitelist. Content acquisition may refer to obtaining data to be detected, and the data to be detected may include source language text, translation corresponding to the source language text, and user interface information of the source language text. The machine translation model is used to perform back translation of the translation. The whitelist refers to a list constructed based on content that is manually confirmed to be a false positive of the detection tool. By continuously updating the whitelist, the detection accuracy of the user interface terminology detection tool 101 can be improved.

[0143] The similarity calculation screening layer may include a bilingual dictionary, a bilingual term list, similarity calculation and scoring. The bilingual dictionary and the bilingual term list are used to construct a back-translation word set. The user interface term detection tool 101 calculates the similarity between the back-translation result and the source language text, and scores based on the similarity calculation result, and screens text pairs (back-translation result and source language text) with scores below a threshold.

[0144] The translation anomaly identification layer may include word frequency statistics, synonyms and translation anomaly segmentation identification. Word frequency statistics may refer to word segmentation and word frequency statistics for the back-translation result and the source language text respectively. The user interface terminology detection tool 101 may identify translation anomaly segmentation from the source language text based on the word frequency statistics of the back-translation result, the word frequency statistics of the source language text and the synonyms.

[0145] The translation necessity assessment layer may include part-of-speech rules, grammar rules, target language style, user interface themes, application information, and wordnet. The translation necessity assessment layer supports filtering abnormal word segmentation of the first translation through part-of-speech rules, scenario association rules, and target language style rules to reduce the number of false positives. The execution order may be part-of-speech rules->scenario association rules->target language style rules. Grammar rules (grammar rules of the target language) and target language style can be used for filtering scenarios of target language style rules, and user interface themes, application information, and wordnet can be used for filtering scenarios of scenario association rules.

[0146] See also Fig. 9 As shown, it is a flow chart of the translation detection method provided in the embodiment of the present application. The translation detection method can be applied to electronic devices, and the electronic devices can be computers, servers and other devices. The translation detection method can include the following steps:

[0147] S901: Acquire data to be detected, where the data to be detected includes a source language text in a user interface and a translation corresponding to the source language text.

[0148] The amount of data to be detected can be determined according to actual needs, and the embodiments of the present application do not limit this. For example, the source language text in the data to be detected can refer to the source language text that appears in all user interfaces in the user device and needs to be translated, and the translation corresponding to the source language text can be the translation of all languages ​​supported by the user device, and the translation can be obtained by machine translation model or manual translation.

[0149] Among the user interface terms of existing user devices, the user interface terms in Chinese and English are relatively mature, and there are basically no problems of missed or wrong translations. That is, the language of the source language text can be set to Chinese or English, and the electronic device can be mainly used to detect whether there are translation anomalies in the translation of the source language text into other languages.

[0150] In some embodiments, since English has similarities with many other languages, setting the source language text to English can further improve the efficiency and accuracy of translation detection and reduce the workload of translation detection.

[0151] In some embodiments, the data to be detected may also include user interface information of the source language text. The user interface information of the source language text may include information of the user interface where the source language text is located and information of the application to which the user interface belongs. For example, the information of the user interface may include information such as the usage scenario and theme of the user interface, and the information of the application to which the user interface belongs may include information such as the type and name of the application.

[0152] In some embodiments, the data to be detected can be pre-stored in a designated storage area, which can be located in the memory of the electronic device itself or in the memory of other electronic devices (such as a server that can communicate with the electronic device). The electronic device can obtain the data to be detected by reading the data to be detected from the designated storage area or receiving the data to be detected sent by other electronic devices.

[0153] S902: Back-translate the translation to obtain a back-translation result corresponding to the translation.

[0154] In some embodiments, the language of the back-translation result is the same as the language of the source language text, and the electronic device can call the translation model to back-translate the translation to obtain a back-translation result corresponding to the translation. The translation model can be pre-installed in the electronic device, and the electronic device can also call the translation model located on the cloud server by accessing the network. The translation model can be an existing machine translation model or a machine translation model developed by the user device manufacturer, which is not limited in the embodiments of the present application.

[0155] In some embodiments, in order to improve the accuracy of the back-translation result, step S902 may specifically include: a1. back-translating the translation to obtain a first back-translation result; a 2. Performing word segmentation processing on the translation to obtain at least one word segment, and searching a bilingual dictionary and / or a bilingual term list to obtain a back-translated word corresponding to each word segment; a 3. Based on the first back-translation result and the found back-translation words, a back-translation result corresponding to the translation is constructed. That is, the back-translation result corresponding to the translation may include one or more back-translation texts corresponding to the translation, which can reduce the back-translation accuracy loss that may be caused by back-translating the translation using a machine translation model.

[0156] S903: Performing word segmentation processing on the back-translation result to obtain a first word segmentation, and performing word segmentation processing on the source language text to obtain a second word segmentation.

[0157] In some embodiments, the electronic device may use an existing word segmentation algorithm to perform word segmentation processing on the back-translation result to obtain one or more first word segmentations, and perform word segmentation processing on the source language text to obtain one or more second word segmentations. The embodiment of the present application does not limit the word segmentation algorithm.

[0158] S904: Determine a first abnormally translated word segment in the second word segment based on the first word segment.

[0159] In some embodiments, the electronic device may compare the first segmented word with the second segmented word to determine the first abnormally translated segmented word in the second segmented word, that is, determine the segmented word that appears in the second segmented word and is suspected of being omitted or mistranslated.

[0160] In some embodiments, the electronic device may use an existing word frequency statistics algorithm to perform word frequency statistics on the first word segmentation to obtain a first word frequency result, perform word frequency statistics on the second word segmentation to obtain a second word frequency result, and determine the first abnormal translation word segmentation in the second word segmentation based on the first word frequency result and the second word frequency result. The embodiment of the present application does not limit the word frequency statistics algorithm.

[0161] For example, if the back-translation result includes a back-translated text, that is, the first word frequency result is a group of word frequency results, the first word frequency result is compared with the second word frequency result. For each word segment in the source language text, if it does not appear in the first word frequency result, or the word frequency number is different from the word frequency number recorded in the first word frequency result, the word segment can be identified as the first translation abnormal word segment.

[0162] For example, if the back-translation result includes multiple back-translated texts, that is, the first word frequency result includes multiple sub-word frequency results, each sub-word frequency result can be compared with the second word frequency result respectively. For each word segment in the source language text, if it does not appear in all sub-word frequency results, or the word frequency number is different from the word frequency number recorded in all sub-word frequency results, the word segment is identified as an abnormal translation word segment.

[0163] In some embodiments, in order to reduce the number of false positives of abnormal translation segmentation, a synonym dictionary may be introduced for retrieval, that is, the electronic device may also determine the first abnormal translation segmentation in the source language text based on the first word frequency result, the second word frequency result and the synonym dictionary. For example, for each segmentation in the source language text, if it does not appear in the first word frequency result, it is determined whether the synonym of the segmentation (obtained based on the synonym dictionary search) appears in the first word frequency result. If the synonym of the segmentation does not appear in the first word frequency result, the segmentation may be identified as the first abnormal translation segmentation.

[0164] S905: Screening the first abnormal translation segmentation words based on the translation necessity evaluation rule, and outputting the screened second abnormal translation segmentation words and the source language text corresponding to the second abnormal translation segmentation words.

[0165] In some embodiments, the translation necessity evaluation rule may include at least one of a part-of-speech rule, a scene association rule, and a target language style rule. The electronic device may score the first translation abnormal participle based on the translation necessity evaluation rule, and filter out the second translation abnormal participle according to the scoring results of each first translation abnormal participle. For example, the electronic device may determine the first translation abnormal participle whose scoring result is greater than a preset score as the second translation abnormal participle.

[0166] Taking the translation necessity evaluation rules including part-of-speech rules, scene association rules and target language style rules as an example, the execution order may be part-of-speech rules->scene association rules->target language style rules. The electronic device screening the first abnormal translation segmentation based on the translation necessity evaluation rules may specifically include: b1. scoring and screening the first abnormal translation segmentation based on the part-of-speech rules to obtain a first screening result; b2. scoring and screening the segmentation in the first screening result based on the scene association rules to obtain a second screening result; b3. scoring and screening the segmentation in the second screening result based on the target language style rules to obtain a second abnormal translation segmentation.

[0167] In some embodiments, the electronic device may also select to execute some or all of the rules in the translation necessity assessment rule according to a preset threshold value of the number of abnormal second translation segmentations. For example, the electronic device scores and screens the first abnormal translation segmentations by part-of-speech rules to obtain a first screening result. If the number of segmentations in the first screening result is less than the preset threshold value, the electronic device ends the scoring and screening, and uses the segmentations in the first screening result as the second abnormal translation segmentations outputted in the end; if the number of segmentations in the first screening result is greater than or equal to the preset threshold value, the electronic device continues to score and screen the segmentations in the first screening result based on the scene association rule to obtain the second screening result. If the number of segmentations in the second screening result is less than the preset threshold value, the electronic device ends the scoring and screening, and uses the segmentations in the second screening result as the second abnormal translation segmentations outputted in the end; if the number of segmentations in the second screening result is still greater than or equal to the preset threshold value, the electronic device continues to score and screen the segmentations in the second screening result based on the target language style rule, and the segmentations obtained by screening are used as the second abnormal translation segmentations outputted in the end.

[0168] User interface terms are mostly short sentences with relatively simple sentence structures. Different scoring weights can be assigned to different parts of speech according to the importance of the parts of speech of each word in the source language text to the semantics. Words can be divided into content words and function words. In the part-of-speech rules, if the source language text includes both content words and function words, the scoring weight of content words can be set to be greater than that of function words. For example, content words can include nouns, verbs, adjectives, numerals, quantifiers, pronouns, etc., and function words can include articles, conjunctions, prepositions, etc.

[0169] If the source language text includes content words of multiple parts of speech, or function words of multiple parts of speech. For various parts of speech in content words or function words, the setting of scoring weights can also be continued. For example, in content words, the scoring weight of nouns is greater than the scoring weight of verbs, and the scoring weight of verbs is greater than the scoring weight of adjectives. In function words, the scoring weight of articles is greater than the scoring weight of conjunctions, and the scoring weight of conjunctions is greater than the scoring weight of prepositions. It can be understood that the scoring weights of various part of speech types in content words can be set according to actual needs, and the scoring weights of various parts of speech in function words can be set according to actual needs, and the embodiments of the present application are not limited to this.

[0170] In some embodiments, when scoring the first translated abnormal word based on the part-of-speech rule, the word frequency of the first translated abnormal word may also be considered. The greater the difference between the word frequency of the first translated abnormal word and the word frequency of the back-translated text, the greater the scoring weight of the first translated abnormal word.

[0171] In some embodiments, the scene association rules can be formulated based on the user interface information of the source language text, which can effectively reduce the false positives of the omitted translation of interface terms. For example, the subject word weight of the user interface of the source language text can be set to be greater than the weight of the application name. For the first translation abnormal word segmentation, the electronic device can calculate the correlation between the first translation abnormal word segmentation and the subject word and application name of the user interface respectively. If the first translation abnormal word segmentation has a higher correlation with the subject word / application name of the user interface (indicating a greater probability of being omitted), the first translation abnormal word segmentation has a lower score, and the possibility of being detected as a false alarm and filtered out is higher.

[0172] In some embodiments, when scoring the scene association rules, the electronic device may also query the vocabulary network (Wordnet) to obtain synonyms, upper and lower concept words of the first abnormal translation participle, and then calculate the correlation with the subject words and application names of the user interface respectively. The higher the correlation, the lower the score. For example, the lowest score among the first abnormal translation participle, synonyms, upper concept words, and lower concept words can be used as the final score of the first abnormal translation participle.

[0173] In some embodiments, the target language style rules can be summarized from books related to the linguistics (vocabulary, grammar rules), translation style, etc. of the target language, and the target language style rules can be used to further reduce the false positives of omitted translation of interface terms. For example, the target language style rules can be constructed based on three aspects: vocabulary, grammar, and N-gram, which can delete words that do not conform to the target language style (thinking habits, language habits, and expressions) to avoid redundant translations, that is, these words that do not conform to the target language style can be omitted and do not constitute translation errors.

[0174] For example, for words that do not conform to the target language style, a lower score is given to facilitate filtering. For example, the target language style rule is a target language style set, which stores multiple style rules related to the target language, and each style rule can be related to a word. When scoring the target language style rule, the score can be based on whether the first translation abnormal word segmentation has appeared in the target language style set. If it has appeared, a relatively low score can be given, and if it has not appeared, a relatively high score can be given.

[0175] In certain embodiments, the second translation abnormal participle output by electronic equipment and the source language text corresponding to the second translation abnormal participle, can be subsequently checked by manual (translator) whether there is indeed translation abnormality. If there is a participle of false alarm in the second translation abnormal participle manually, the participle of these false alarms and the corresponding source language text can be added to the white list, so that subsequent electronic equipment can improve the filtering ability of the false alarm participle based on the updated white list. The content in the white list can be updated according to the manual inspection result, that is, the electronic equipment can be based on the translation necessity assessment rule and the regularly updated white list (white list can also be integrated into the translation necessity assessment rule) to screen and obtain the second translation abnormal participle.

[0176] See also Fig.10 As shown, it is a flow chart of the translation detection method provided in the embodiment of the present application. The translation detection method can be applied to electronic devices, and the electronic devices can be computers, servers and other devices. The translation detection method can include the following steps:

[0177] S101: Acquire data to be detected, where the data to be detected includes a source language text in a user interface and a translation corresponding to the source language text.

[0178] Step S101 of the embodiment of the present application is similar to step S901 of the aforementioned embodiment, and will not be described again here to avoid repetition.

[0179] S102: Back-translate the translation to obtain a back-translation result corresponding to the translation.

[0180] Step S102 of the embodiment of the present application is similar to step S902 of the aforementioned embodiment, and will not be described again here to avoid repetition.

[0181] S103: Calculate the similarity between the source language text and the corresponding back-translation result, and add the source language text suspected of abnormal translation and the corresponding back-translation result to the text set based on the calculated similarity.

[0182] In some embodiments, the electronic device may use an existing text similarity calculation algorithm to calculate the similarity between the source language text and the corresponding back-translation result. The embodiments of the present application do not limit the text similarity calculation algorithm.

[0183] In some embodiments, the similarity between the back-translation result and the corresponding source language text may also refer to semantic similarity, and the electronic device may also calculate the semantic similarity between the source language text and the corresponding back-translation result based on a semantic similarity calculation algorithm. The present application embodiment does not limit the semantic similarity calculation algorithm.

[0184] If the source language text and the corresponding back-translation result include multiple back-translation texts, the electronic device may also calculate the similarity between each back-translation text and the source language text, and use the maximum similarity among the multiple similarities calculated as the similarity between the source language text and the corresponding back-translation result. In other embodiments, the electronic device may also use the average or median value of the multiple similarities as the similarity between the source language text and the corresponding back-translation result.

[0185] Since the back-translation result is associated with the translation, the similarity between the source language text and the corresponding back-translation result can be used to characterize the translation accuracy of the translation. The higher the similarity between the source language text and the corresponding back-translation result, the higher the accuracy of the translation obtained based on the source language text. Conversely, the lower the similarity between the source language text and the corresponding back-translation result, the lower the accuracy of the translation obtained based on the source language text.

[0186] In some implementations, a similarity threshold can be set in advance, and the calculated similarity can be compared with the similarity threshold to screen out source language texts with suspected translation anomalies and the corresponding back-translation results. If the similarity between a source language text and the corresponding back-translation result is greater than or equal to the similarity threshold, it indicates that the accuracy of the translation obtained based on the source language text translation meets the requirements, and it can be considered that there are no translation errors (omissions or mistranslations) in the translation, and there is no need to process the back-translation results of the translation and the source language text corresponding to the translation. If the similarity between a source language text and the corresponding back-translation result is less than the similarity threshold, it indicates that the accuracy of the translation obtained based on the source language text translation does not meet the requirements, and it can be considered that there are suspected translation errors (omissions or mistranslations) in the translation. The source language text and the corresponding back-translation result can be added to the text set with suspected translation anomalies to continue analysis and processing based on subsequent steps.

[0187] Since user interface terms are mostly short sentences with low literary value, the translation accuracy of translations obtained based on machine translation models or human translators is relatively high, and there are only a few user interface terms with translation anomalies. Electronic devices can reduce data processing volume, save computing resources, and shorten the detection path of these texts without anomalies by filtering source language texts and corresponding back-translation results with a similarity greater than or equal to a similarity threshold.

[0188] S104: performing word segmentation processing on the back-translation results in the text set to obtain first word segmentations, and performing word segmentation processing on the source language text corresponding to the back-translation results in the text set to obtain second word segmentations.

[0189] In some embodiments, the electronic device may use an existing word segmentation algorithm to perform word segmentation processing on the back-translation results in the text set to obtain one or more first word segmentations, and perform word segmentation processing on the source language text in the text set to obtain one or more second word segmentations. The embodiments of the present application do not limit the word segmentation algorithm.

[0190] S105: Determine a first abnormally translated word segment in the second word segment based on the first word segment.

[0191] Step S105 of the embodiment of the present application is similar to step S904 of the aforementioned embodiment, and will not be described again here to avoid repetition.

[0192] S106: Screening the first abnormal translation segmentation words based on the translation necessity evaluation rule, and outputting the screened second abnormal translation segmentation words and the source language text corresponding to the second abnormal translation segmentation words.

[0193] Step S106 of the embodiment of the present application is similar to step S905 of the aforementioned embodiment, and will not be described again here to avoid repetition.

[0194] like Fig.11 As shown, the electronic device 100 may include a processor 1001, a memory 1002, and a communication bus 1003. The memory 1002 is used to store one or more computer programs 1004. The one or more computer programs 1004 are configured to be executed by the processor 1001. The one or more computer programs 1004 include instructions, and the processor 1001 executes the above instructions to implement the following in the electronic device 100: Figure 5 , Fig. 9 or Fig.10 The electronic device 100 may be a computer, a server or other device.

[0195] It is understandable that the structure shown in this embodiment does not constitute a specific limitation on the electronic device 100. In other embodiments, the electronic device 100 may include more or fewer components than shown, or combine or separate some components, or arrange the components differently.

[0196] The processor 1001 may include one or more processing units, for example, the processor 1001 may include an application processor (AP), a modem, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), etc. Different processing units may be independent devices or integrated into one or more processors.

[0197] The processor 1001 may also be provided with a memory for storing instructions and data. In some embodiments, the memory in the processor 1001 is a cache memory. The memory may store instructions or data that the processor 1001 has just used or cyclically used. If the processor 1001 needs to use the instruction or data again, it may be directly called from the memory. This avoids repeated access, reduces the waiting time of the processor 1001, and thus improves the efficiency of the system.

[0198] In some embodiments, the processor 1001 may include one or more interfaces. The interface may include an inter-integrated circuit (I2C) interface, an inter-integrated circuit sound (I2S) interface, a pulse code modulation (PCM) interface, a universal asynchronous receiver / transmitter (UART) interface, a mobile industry processor interface (MIPI), a general-purpose input / output (GPIO) interface, a SIM interface, and / or a USB interface, etc.

[0199] In some embodiments, memory 1002 may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0200] See also Fig.12 As shown, the user device 200 can be a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a notebook computer, an ultra mobile personal computer (UMPC), a netbook, a cellular phone, a personal digital assistant (PDA), an augmented reality (AR) device, a virtual reality (VR) device, an artificial intelligence (AI) device, a wearable device, a vehicle-mounted device, a smart home device and / or a smart city device. The embodiment of the present application does not impose any special restrictions on the specific type of the user device 200.

[0201] The user device 200 may include a processor 110, an external memory interface 120, an internal memory 121, a Universal Serial Bus (USB) interface 130, a charging management module 140, a power management module 141, a battery 142, an antenna 1, an antenna 2, a mobile communication module 150, a wireless communication module 160, an audio module 170, a speaker 170A, a receiver 170B, a microphone 170C, an earphone interface 170D, a sensor module 180, a button 190, a motor 191, an indicator 192, a camera 193, a display screen 194, and a Subscriber Identification Module (SIM) card interface 195, etc. The sensor module 180 may include a pressure sensor 180A, a gyroscope sensor 180B, an air pressure sensor 180C, a magnetic sensor 180D, an acceleration sensor 180E, a distance sensor 180F, a proximity light sensor 180G, a fingerprint sensor 180H, a temperature sensor 180J, a touch sensor 180K, an ambient light sensor 180L, a bone conduction sensor 180M, etc.

[0202] It is to be understood that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the user equipment 200. In other embodiments of the present application, the user equipment 200 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0203] The processor 110 may include one or more processing units, for example, the processor 110 may include an application processor (AP), a modem processor, a graphics processor (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), etc. Different processing units may be independent devices or integrated into one or more processors.

[0204] The controller can generate operation control signals according to the instruction operation code and timing signal to complete the control of instruction fetching and execution.

[0205] The processor 110 may also be provided with a memory for storing instructions and data. In some embodiments, the memory in the processor 110 is a cache memory. The memory may store instructions or data that the processor 110 has just used or cyclically used. If the processor 110 needs to use the instruction or data again, it may be directly called from the memory. This avoids repeated access, reduces the waiting time of the processor 110, and thus improves the efficiency of the system.

[0206] In some embodiments, the processor 110 may include one or more interfaces. The interface may include an Inter-integrated Circuit (I2C) interface, an Inter-integrated Circuit Sound (I2S) interface, a Pulse Code Modulation (PCM) interface, a Universal Asynchronous Receiver / Transmitter (UART) interface, a Mobile Industry Processor Interface (MIPI), a General-Purpose Input / Output (GPIO) interface, a Subscriber Identity Module (SIM) interface, and / or a Universal Serial Bus (USB) interface, etc.

[0207] The I2C interface is a bidirectional synchronous serial bus, including a serial data line (SDA) and a serial clock line (SCL). In some embodiments, the processor 110 may include multiple groups of I2C buses. The processor 110 may be coupled to the touch sensor 180K, the charger, the flash, the camera 193, etc. through different I2C bus interfaces. For example: the processor 110 may be coupled to the touch sensor 180K through the I2C interface, so that the processor 110 communicates with the touch sensor 180K through the I2C bus interface to realize the touch function of the user device 200.

[0208] The I2S interface can be used for audio communication.

[0209] The PCM interface can also be used for audio communication to sample, quantize and encode analog signals.

[0210] The UART interface is a universal serial data bus used for asynchronous communication. The bus can be a bidirectional communication bus. It converts the data to be transmitted between serial communication and parallel communication.

[0211] The MIPI interface can be used to connect the processor 110 with peripheral devices such as the display screen 194 and the camera 193. The MIPI interface includes a camera serial interface (CSI), a display serial interface (DSI), etc. In some embodiments, the processor 110 and the display screen 194 communicate via the DSI interface to implement the display function of the user device 200.

[0212] The GPIO interface can be configured by software. The GPIO interface can be configured as a control signal or a data signal.

[0213] The USB interface 130 is an interface that complies with USB standard specifications, and may specifically be a Mini USB interface, a Micro USB interface, a USB Type C interface, etc.

[0214] It is understandable that the interface connection relationship between the modules illustrated in the embodiment of the present application is only a schematic illustration and does not constitute a structural limitation on the user equipment 200. In other embodiments of the present application, the user equipment 200 may also adopt different interface connection methods in the above embodiments, or a combination of multiple interface connection methods.

[0215] The user device 200 implements display functions through a GPU, a display screen 194, and an application processor. The GPU is a microprocessor for image processing, connected to the display screen 194 and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. The processor 110 may include one or more GPUs that execute program instructions to generate or change display information.

[0216] The display screen 194 is used to display images, videos, etc. The display screen 194 includes a display panel. The display panel can be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode or an active-matrix organic light-emitting diode (AMOLED), a flexible light-emitting diode (FLED), Miniled, Microled, Micro-OLED, a quantum dot light-emitting diode (QLED), etc. In some embodiments, the user device 200 may include 1 or N display screens 194, where N is a positive integer greater than 1.

[0217] The digital signal processor is used to process digital signals, and can process not only digital image signals but also other digital signals. For example, when the user equipment 200 is selecting a frequency point, the digital signal processor is used to perform Fourier transform on the frequency point energy.

[0218] The internal memory 121 may include one or more random-access memories (RAM) and one or more non-volatile memories (NVM).

[0219] Random access memory may include static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM, for example, the fifth generation DDR SDRAM is generally referred to as DDR5 SDRAM), etc.; non-volatile memory may include disk storage devices and flash memory.

[0220] Flash memory can be divided into NOR FLASH, NAND FLASH, 3D NAND FLASH, etc. according to the operating principle. It can be divided into single-level cell (SLC), multi-level cell (MLC), triple-level cell (TLC), quad-level cell (QLC), etc. according to the storage specification. It can be divided into universal flash storage (UFS), embedded Multi Media Card (eMMC), etc. according to the storage specification.

[0221] The random access memory can be directly read and written by the processor 110, and can be used to store executable programs (such as machine instructions) of the operating system or other running programs, and can also be used to store user and application data.

[0222] The non-volatile memory may also store executable programs and user and application data, etc., and may be loaded into the random access memory in advance for direct reading and writing by the processor 110 .

[0223] The external memory interface 120 can be used to connect to an external non-volatile memory to expand the storage capacity of the user device 200. The external non-volatile memory communicates with the processor 110 through the external memory interface 120 to implement a data storage function. For example, files such as music and videos are stored in the external non-volatile memory.

[0224] The internal memory 121 or the external memory interface 120 is used to store one or more computer programs. The one or more computer programs are configured to be executed by the processor 110 .

[0225] The pressure sensor 180A is used to sense the pressure signal and can convert the pressure signal into an electrical signal. In some embodiments, the pressure sensor 180A can be set on the display screen 194. There are many types of pressure sensors 180A, such as resistive pressure sensors, inductive pressure sensors, capacitive pressure sensors, etc. The capacitive pressure sensor can be a parallel plate including at least two conductive materials. When a force acts on the pressure sensor 180A, the capacitance between the electrodes changes. The user device 200 determines the intensity of the pressure according to the change in capacitance. When a touch operation acts on the display screen 194, the user device 200 detects the touch operation intensity according to the pressure sensor 180A. The user device 200 can also calculate the touch position according to the detection signal of the pressure sensor 180A. In some embodiments, touch operations acting on the same touch position but with different touch operation intensities can correspond to different operation instructions. For example: when a touch operation with a touch operation intensity less than the first pressure threshold acts on the short message application icon, an instruction to view the short message is executed. When a touch operation with a touch operation intensity greater than or equal to the first pressure threshold acts on the short message application icon, an instruction to create a new short message is executed.

[0226] The touch sensor 180K is also called a "touch control device". The touch sensor 180K can be set on the display screen 194, and the touch sensor 180K and the display screen 194 form a touch screen, also called a "touch control screen". The touch sensor 180K is used to detect touch operations acting on or near it. The touch sensor can pass the detected touch operation to the application processor to determine the type of touch event. Visual output related to the touch operation can be provided through the display screen 194. In other embodiments, the touch sensor 180K can also be set on the surface of the user device 200, which is different from the position of the display screen 194.

[0227] The bone conduction sensor 180M can obtain a vibration signal. In some embodiments, the bone conduction sensor 180M can obtain a vibration signal of a vibrating bone block of the vocal part of the human body. The bone conduction sensor 180M can also contact the human pulse to receive a blood pressure beat signal. In some embodiments, the bone conduction sensor 180M can also be set in an earphone and combined into a bone conduction earphone. The audio module 170 can parse out a voice signal based on the vibration signal of the vibrating bone block of the vocal part obtained by the bone conduction sensor 180M to realize a voice function. The application processor can parse the heart rate information based on the blood pressure beat signal obtained by the bone conduction sensor 180M to realize a heart rate detection function.

[0228] The button 190 includes a power button, a volume button, etc. The button 190 may be a mechanical button or a touch button. The user device 200 may receive a button input and generate a key signal input related to the user settings and function control of the user device 200.

[0229] Motor 191 can generate vibration prompts. Motor 191 can be used for incoming call vibration prompts, and can also be used for touch vibration feedback. For example, touch operations acting on different applications (such as taking pictures, audio playback, etc.) can correspond to different vibration feedback effects. For touch operations acting on different areas of the display screen 194, motor 191 can also correspond to different vibration feedback effects. Different application scenarios (for example: time reminders, receiving messages, alarm clocks, games, etc.) can also correspond to different vibration feedback effects. The touch vibration feedback effect can also support customization.

[0230] The indicator 192 may be an indicator light, which may be used to indicate the charging status, power changes, messages, missed calls, notifications, etc.

[0231] The SIM card interface 195 is used to connect a SIM card. The SIM card can be connected to and separated from the user device 200 by inserting it into the SIM card interface 195 or pulling it out from the SIM card interface 195. The user device 200 can support 1 or N SIM card interfaces, where N is a positive integer greater than 1. The SIM card interface 195 can support Nano SIM cards, Micro SIM cards, SIM cards, and the like. Multiple cards can be inserted into the same SIM card interface 195 at the same time. The types of the multiple cards can be the same or different. The SIM card interface 195 can also be compatible with different types of SIM cards. The SIM card interface 195 can also be compatible with external memory cards. The user device 200 interacts with the network through the SIM card to implement functions such as calls and data communications. In some embodiments, the user device 200 uses an eSIM, i.e., an embedded SIM card. The eSIM card can be embedded in the user device 200 and cannot be separated from the user device 200.

[0232] This embodiment further provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on the electronic device 100, the electronic device 100 executes the above-mentioned related method steps to implement the translation detection method in the above-mentioned embodiment.

[0233] This embodiment further provides a computer program product. When the computer program product is run on a computer, the computer is caused to execute the above-mentioned related steps to implement the translation detection method in the above-mentioned embodiment.

[0234] In addition, an embodiment of the present application also provides a device, which may specifically be a chip, a component or a module, and the device may include a connected processor and a memory; wherein the memory is used to store computer-executable instructions, and when the device is running, the processor may execute the computer-executable instructions stored in the memory, so that the chip executes the translation detection method in the above-mentioned method embodiments.

[0235] Among them, the electronic device, computer-readable storage medium, computer program product or chip provided in this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above and will not be repeated here.

[0236] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0237] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0238] The unit described as a separate component may or may not be physically separated, and the component shown as a unit may be one physical unit or multiple physical units, that is, it may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.

[0239] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0240] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to perform all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random-Access Memory, RAM), disk or optical disk and other media that can store program code.

[0241] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.< / canvas> < / video> < / videoview> < / imgview> < / textview>

Claims

1. A translation detection method, characterized in that: The method comprises: Acquire data to be detected, wherein the data to be detected includes a source language text in a user interface, a translation corresponding to the source language text, and user interface information of the source language text, wherein the user interface information includes a theme of the user interface and / or a name of an application to which the user interface belongs; Back-translating the translation to obtain a back-translation result corresponding to the translation; Performing word segmentation processing on the back-translation result to obtain a first word segmentation, and performing word segmentation processing on the source language text to obtain a second word segmentation; Determine a first abnormal translation segmentation in the second segmentation based on the first segmentation; The first abnormal translation participle is screened based on a translation necessity assessment rule, and the screened second abnormal translation participle and the source language text corresponding to the second abnormal translation participle are output, wherein the translation necessity assessment rule includes a scene association rule, and the scene association rule is formulated based on user interface information of the source language text.

2. The translation detection method according to claim 1, characterized in that: The back-translating the translation to obtain a back-translation result corresponding to the translation includes: Back-translating the translation to obtain a first back-translation result; Segmenting the translation to obtain third segmented words, and searching a bilingual dictionary and / or a bilingual term list to obtain a back-translated word corresponding to each third segmented word; Based on the first back-translation result and the back-translation word, a back-translation result corresponding to the translation is obtained, and the back-translation result corresponding to the translation includes one or more back-translation texts corresponding to the translation.

3. The translation detection method according to claim 1, characterized in that: After obtaining the back-translation result corresponding to the translation, the method further includes: The similarity between the source language text and the corresponding back-translation result is calculated, and based on the calculated similarity, the source language text suspected of having an abnormal translation and the corresponding back-translation result are added to the text set.

4. The translation detection method according to claim 3, characterized in that: The performing word segmentation processing on the back-translation result to obtain the first word segmentation, and performing word segmentation processing on the source language text to obtain the second word segmentation, comprises: The back-translated results in the text set are processed by word segmentation to obtain first word segments, and the source language text in the text set corresponding to the back-translated results is processed by word segmentation to obtain second word segments.

5. The translation detection method according to any one of claims 1 to 4, characterized in that: The determining, based on the first segmentation, a first abnormal translation segmentation in the second segmentation includes: Performing word frequency statistics on the first word segmentation to obtain a first word frequency result; Performing word frequency statistics on the second word segmentation to obtain a second word frequency result; Based on the first word frequency result and the second word frequency result, a first abnormally translated word segment in the second word segment is determined.

6. The translation detection method according to claim 5, characterized in that: The determining, based on the first word frequency result and the second word frequency result, a first abnormally translated word segmentation in the second word segmentation includes: Based on the first word frequency result, the second word frequency result and a synonym dictionary, a first abnormally translated word segment in the second word segment is determined, and the synonym dictionary is used to find a synonym of the first word segment.

7. The translation detection method according to any one of claims 1 to 4, characterized in that: The screening of the first abnormal translation segmentation words based on the translation necessity evaluation rule includes: Scoring the first abnormal translation segmentation based on the translation necessity evaluation rule; The first abnormal translation participle whose score result is greater than the preset score is determined as the second abnormal translation participle.

8. The translation detection method according to any one of claims 1 to 4, characterized in that: The translation necessity evaluation rule also includes part-of-speech rules and target language style rules. The screening of the first translation abnormal word segmentations based on the translation necessity evaluation rule includes: Scoring and screening the first abnormal translation segmentations based on the part-of-speech rule to obtain a first screening result; Scoring and filtering the segmented words in the first screening result based on the scene association rule to obtain a second screening result; The word segmentations in the second screening result are scored and screened based on the target language style rule to obtain the second translation abnormal word segmentations.

9. The translation detection method according to claim 8, characterized in that: After obtaining the first screening result, the method further includes: If the number of segmented words in the first screening result is less than a preset number threshold, taking the segmented words in the first screening result as the second abnormal translation segmented words; The screening of the first screening result based on the scene association rule includes: If the number of word segments of the first screening result is greater than or equal to a preset number threshold, the first screening result is screened based on the scene association rule.

10. The translation detection method according to claim 9, characterized in that: After obtaining the second screening result, the method further includes: If the number of segmented words in the second screening result is less than the preset number threshold, taking the segmented words in the second screening result as the second abnormal translation segmented words; The screening of the second screening result based on the target language style rule includes: If the number of word segments of the second screening result is greater than or equal to the preset number threshold, the second screening result is screened based on the target language style rule.

11. The translation detection method according to claim 8, characterized in that: The part-of-speech rule includes: if the source language text includes participles of multiple parts of speech, scoring is performed based on the part-of-speech of the participles, wherein the scoring weight of content words is greater than the scoring weight of function words.

12. The translation detection method according to claim 8, characterized in that: The scene association rule includes: scoring based on the association between the segmentation and the theme of the user interface where the source language text is located, and based on the association between the segmentation and the name of the application to which the user interface belongs, the association is negatively correlated with the scoring score, and the scoring weight of the theme of the user interface is greater than the scoring weight of the name of the application.

13. The translation detection method according to claim 8, characterized in that: The target language style rules include: scoring based on whether the word segmentation conforms to the target language style.

14. The translation detection method according to any one of claims 1 to 4, characterized in that: After outputting the second abnormal translation segmentation obtained by screening and the source language text corresponding to the second abnormal translation segmentation, the method further includes: Obtaining a review result of the second translation abnormal word segmentation, and adding the second translation abnormal word segmentation whose review result is a misdetection and the source language text corresponding to the second translation abnormal word segmentation to a whitelist; The whitelist is added to the translation necessity evaluation rule.

15. An electronic device, characterized in that: The electronic device comprises a memory and a processor: Wherein, the memory is used to store program instructions; The processor is used to read and execute the program instructions stored in the memory. When the program instructions are executed by the processor, the electronic device executes the translation detection method according to any one of claims 1 to 14.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed on an electronic device, the electronic device executes the translation detection method according to any one of claims 1 to 14.

17. A chip coupled to a memory in an electronic device, characterized in that: The chip is used to control the electronic device to execute the translation detection method according to any one of claims 1 to 14.

Images