An embedded translation system and method with content censorship mechanism

Abstract

The embodiment of the present application provides a kind of embedded translation system and method with content auditing mechanism, it is related to the technical field of embedded translation technology.The method comprises: obtaining translation request;Call translation interface to translate the text to be translated to obtain initial text;Call auditing engine to audit the initial text, and in the case where the audit result is that the initial text meets the preset condition, the initial text is used as target text;Interface adaptive matching is carried out according to the target text, to display the target text.By the present application, the problem of low translation accuracy is solved, and the effect of improving translation accuracy and efficiency is achieved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of embedded translation, and in particular, to an embedded translation system and method with a content review mechanism. BACKGROUND

[0002] As an important part of modern information technology, embedded translation systems have been deeply integrated into smart phones, wearable devices, vehicle-mounted systems, and various Internet of Things terminals, providing great convenience for cross-language communication. However, while pursuing translation speed and fluency, existing embedded translation technology often overlooks the depth control of output content quality. The results of machine translation are usually presented directly to the user or directly used for business publishing without review. In informal daily communication scenarios, this mode is still acceptable, but in serious business scenarios involving professional terms, legal provisions, marketing, brand promotion, etc., its inherent defects are exposed.

[0003] Specifically, 1. Professional terms are not unified. For the same professional concept, machine translation may give different translations in different contexts, which damages the rigor and consistency of professional documents. 2. The content is not accurate. The translation result may misinterpret the precise meaning of the original text, leading to distorted information transmission, misunderstandings, and even business risks. 3. The content is not compliant. The translation process may generate or retain content that violates relevant laws and regulations (such as limit words in the Advertising Law, data privacy clauses), bringing legal risks to enterprises.

[0004] Existing solutions either rely on manual post-editing, which is inefficient and costly, and cannot meet the real-time requirements of embedded scenarios, or use simple keyword filtering, which is difficult to address context-related, deep compliance and professionalism issues. Therefore, how to build a translation system that can embed a mandatory review link in a resource-constrained embedded environment to effectively improve the accuracy, professionalism, and compliance of translated content and reduce business risks is an important technical challenge in the field. SUMMARY

[0005] Embodiments of the present application provide an embedded translation system and method with a content review mechanism to at least solve the problem of inaccurate translation in related technologies.

[0006] According to an embodiment of the present application, an embedded translation method with a content review mechanism is provided, comprising:

[0007] Obtaining a translation request, wherein the translation request is used to indicate a request to translate a first language text to be translated into a second language target text;

[0008] Calling a translation interface to translate the text to be translated to obtain an initial text;

[0009] calling an audit engine to audit the initial text, and in a case where an audit result is that the initial text meets preset conditions, taking the initial text as a target text;

[0010] performing interface self-adaptive matching according to the target text to display the target text.

[0011] In an example embodiment, the calling a translation interface to translate the text to be translated to obtain an initial text comprises:

[0012] obtaining text to be translated, and constructing a structured semantic representation based on the text to be translated;

[0013] performing semantic mapping processing on the structured semantic representation to generate a candidate translation set containing potential semantic flaw units;

[0014] performing risk aggregation evolution processing on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set;

[0015] determining the initial text based on the aggregated risk semantic clusters.

[0016] In an example embodiment, the performing risk aggregation evolution processing on the candidate translation set comprises:

[0017] calculating an initial risk potential value of each semantic flaw unit in the candidate translation set based on a preset knowledge base;

[0018] iteratively updating the risk potential value according to a preset iterative diffusion rule to obtain a target risk potential value;

[0019] clustering semantic flaw units whose target risk potential values exceed a preset threshold to obtain the aggregated risk semantic clusters.

[0020] In an example embodiment, the calling an audit engine to audit the initial text comprises:

[0021] calling the audit engine to determine a flaw type and a flaw level of the aggregated risk semantic clusters;

[0022] performing a preset purification strategy on text content corresponding to the aggregated risk semantic clusters according to the flaw type and the flaw level, wherein the purification strategy comprises uniform inconsistent term translation, replacement of non-compliant text content, or interception of unsafe text content;

[0023] Integrate the text content after the purification strategy processing and meeting the preset conditions with the text content of the non-flawed part in the candidate translation set, and generate the target text.

[0024] In one example embodiment, the interface adaptive matching according to the target text comprises:

[0025] Obtaining length information of the target text;

[0026] Based on the length information, dynamically adjusting the size or layout of the display area in the user interface for displaying the target text.

[0027] According to another embodiment of the present application, an embedded translation system with a content review mechanism is provided, comprising:

[0028] A request acquisition module is configured to obtain a translation request, wherein the translation request is used to indicate a request to translate a to-be-translated text in a first language into a target text in a second language;

[0029] A translation module is configured to call a translation interface to translate the to-be-translated text to obtain an initial text;

[0030] A text review module is configured to call a review engine to review the initial text, and in a case where the review result is that the initial text meets a preset condition, the initial text is taken as a target text;

[0031] An interface matching module is configured to perform interface adaptive matching according to the target text to display the target text.

[0032] In one example embodiment, the calling of the translation interface to translate the to-be-translated text to obtain an initial text comprises:

[0033] Obtaining a to-be-translated text, and constructing a structured semantic representation based on the to-be-translated text;

[0034] Performing semantic mapping processing on the structured semantic representation to generate a candidate translation set containing potential semantic flaw units;

[0035] Performing risk aggregation evolution processing on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set;

[0036] Determining the initial text based on the aggregated risk semantic clusters.

[0037] In one example embodiment, the performing of risk aggregation evolution processing on the candidate translation set comprises:

[0038] calculate an initial risk potential value of each semantic flaw unit in the candidate translation set based on a preset knowledge base;

[0039] update the risk potential value iteratively according to a preset iterative diffusion rule to obtain a target risk potential value;

[0040] cluster the semantic flaw units whose target risk potential values exceed a preset threshold to obtain the aggregated risk semantic cluster.

[0041] According to still another embodiment of the present application, there is also provided a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when executed.

[0042] According to still another embodiment of the present application, there is also provided an electronic device comprising a memory and a processor, the memory having a computer program stored therein, and the processor being arranged to execute the computer program to perform the steps of any of the method embodiments described above.

[0043] By the present application, since accurate translation is performed by calling a translation interface, content review is performed by a review engine, and text content that does not meet requirements is adjusted, the translation accuracy is effectively improved, so that the problem of low translation accuracy can be solved, and the effect of improving translation accuracy and efficiency is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 is a flowchart of an embedded translation method with a content review mechanism according to an embodiment of the present application;

[0045] Figure 2 is a structural block diagram of an embedded translation system with a content review mechanism according to an embodiment of the present application. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0047] Hereinafter, the terms "first", "second", etc. are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", etc. can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0048] In addition, in the present application, the orientation terms such as "upper", "lower", "left", "right" and the like can include, but are not limited to, the orientation defined by the relative placement of the components in the drawings. It should be understood that these directional terms are relative concepts, which are used for relative description and clarification, and can change accordingly according to the change of the placement of the components in the drawings.

[0049] In the present application, unless otherwise explicitly specified and limited, the term "connection" should be understood broadly, for example, "connection" can be fixed connection, or detachable connection, or integral; can be directly connected, or indirectly connected through an intermediate medium. In addition, the term "coupling" can be an electrically connected manner for signal transmission.

[0050] As used herein, "about", "approximately", or "around" includes the stated value and the average value within an acceptable deviation range of the specific value, wherein the acceptable deviation range is determined by the person of ordinary skill in the art considering the measurement being discussed and the error related to the measurement of the specific quantity (i.e., the limitation of the measurement system).

[0051] The technical concept is to provide an intelligent and deep review and purification layer for an external AI translation interface. Instead of relying on the single best result returned by the AI translation interface, the method actively obtains multiple candidate translations and constructs them into an unstable intermediate data structure, "metastable candidate translation set". In this structure, various potential content quality problems (such as inconsistency in professional term translation, expressions that violate regulations, etc.) generated by general AI models are defined as "semantic flaw units". The key innovation of the present application is that through a controlled "risk aggregation evolution process" calculation process, these scattered and hidden "semantic flaw units" are spontaneously aggregated between multiple candidate translations or within a single translation, forming "aggregated risk semantic clusters" that are structurally visible and semantically concentrated. Once these semantic clusters are formed, the system can perform highly efficient and accurate positioning, analysis and repair, and finally output a high-standard translation that far exceeds the quality of the original AI interface.

[0052] In the present embodiment, an embedded translation method with a content review mechanism is provided, Figure 1 is a flowchart of an embedded translation method with a content review mechanism according to an embodiment of the present application, as shown in Figure 1 The flowchart includes the following steps:

[0053] In step S11, a translation request is obtained, wherein the translation request is used to indicate a request to translate a text to be translated in a first language into a target text in a second language;

[0054] In the embodiment, when the user needs to translate the text, the user uploads the text to be translated to the target area, selects the language to be translated, and then clicks the "translation button" to generate a translation request; for example, Chinese to English, or Chinese to Japanese, etc.

[0055] Step S12, calling a translation interface to translate the text to be translated to obtain an initial text;

[0056] In the embodiment, the called translation interface can be an AI interface, and when the AI interface is used, the keywords of the text to be translated can be recognized, and the obtained keywords are input to the corresponding translation model through the AI interface to improve the translation accuracy.

[0057] The calling of the translation interface to translate the text to be translated to obtain an initial text includes:

[0058] Step S121, obtaining the text to be translated, and constructing a structured semantic representation based on the text to be translated;

[0059] Step S122, performing semantic mapping processing on the structured semantic representation to generate a candidate translation set containing potential semantic flaw units;

[0060] Step S123, performing risk aggregation evolution processing on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set;

[0061] Step S124, determining the initial text based on the aggregated risk semantic cluster.

[0062] In the embodiment, the system obtains the language text that the user wants to translate through a preset user interface or program interface, and pre-processes it, and then performs risk aggregation to form an initial text.

[0063] For example, the user inputs a text stream through a keyboard, pastes it from a clipboard, or converts it from speech by an ASR module; meanwhile, the user usually specifies the target language of the translation, such as from "Chinese" to "English"; then it is constructed into a structured semantic representation, which contains the syntax structure, word vectors, and other information of the source text, although it is not directly used for translation requests, but can be used as auxiliary information for subsequent processing to improve the accuracy of risk aggregation evolution processing; Specifically, it can first perform lexical analysis, including word segmentation (cutting continuous character strings into meaningful word units) and part-of-speech tagging (assigning each word to its grammatical role, such as noun, verb, adjective, etc.); Then, perform syntactic analysis, the core of which is to construct a dependency graph of the sentence, which takes words as nodes and grammatical dependency relationships as edges, accurately revealing the internal structure of the sentence, such as which word is the subject, which is the predicate, and their modification and modified relationship; Finally, perform deep semantic analysis, which mainly generates context-related word embedding vectors for each word unit, which is a mathematical representation of the semantics of the word in a high-dimensional continuous space, so that words with similar semantics are also close in space. This embodiment can use a lightweight pre-trained language model (such as a variant of MobileBERT or DistilBERT) that is suitable for embedded environments and has been knowledge distilled and quantitatively compressed to efficiently generate these word vectors; Finally, the word vector sequence obtained by the above analysis, the syntactic dependency structure (which can be represented as an adjacency matrix or adjacency list), and some global context feature information (such as the topic vector of the article) are integrated together to form a unified and structured structured semantic representation.

[0064] Then call the AI translation interface for preliminary translation. The translation process here is not performed by the system itself, but the obtained language text is sent to a preset AI translation interface through the network. This interface can be an external interface provided by a third-party service provider, such as a commercial translation API, or an internal model interface.

[0065] For example, the system can set the parameter num_return_sequences=3 or num_beams=5 (the specific parameter name depends on the definition of the API), requesting to return 3 or 5 most likely translation results; after receiving the data returned by the AI translation interface (usually in JSON format, containing a list of candidate translations and their corresponding confidence scores), the system parses the data and constructs a unified data structure, i.e., a candidate translation set, from the N candidate translations.

[0066] For example, the system can set the parameter num_return_sequences=3 or num_beams=5 (the specific parameter name depends on the definition of the API), requesting to return 3 or 5 most likely translation results; after receiving the data returned by the AI translation interface (usually in JSON format, containing a list of candidate translations and their corresponding confidence scores), the system parses the data and constructs a unified data structure, i.e., a candidate translation set, from the N candidate translations.

[0067] Subsequently, the audit engine is called to iteratively calculate the candidate translation set to obtain the aggregated risk semantic cluster, specifically including:

[0068] The audit engine computes an initial risk potential value for each semantic unit (e.g. word, phrase) in all candidate translations based on one or more knowledge bases related to the business scenario, e.g. loading a professional term knowledge base which specifies that the only correct translation for "A project" is "Project A", then "The aProject" and "a project" in candidate translations 2 and 3 will be assigned higher initial risk potential, and so on.

[0069] The audit engine then iteratively updates the risk potential in a pre-defined semantic topology space which not only defines the adjacency relationship of words within a single translation, but also establishes the connection between corresponding words in different candidate translations. The iteration process makes high potential risk points (e.g. non-compliant words) transmit their influence to the surrounding words, while conflicting translations for the same concept (e.g. "Project A" vs "The A Project") in different candidate translations interact with each other to form a strong potential peak in this semantic area; while in the second iteration, the risk potential value of the unit

[0070]

[0071] is the simulated time step, controlling the update rate; is the neighbor set of is the weight coefficient, representing the influence strength of unit on .

[0072] After the iteration ends, the system identifies all semantic units with risk potential exceeding a pre-defined threshold as risk core points, and clusters their neighboring high-risk points of the same type in the semantic topology space to form aggregated risk semantic clusters .

[0073] For example, the system forms two main aggregated risk semantic clusters:

[0074] (Term inconsistency): This cluster aggregates all three different translations for "A project" in all candidate translations to form a high potential area, clearly indicating that there is a term consistency problem here.

[0075] (Compliance risk): This cluster aggregates "best" and "the best" appearing in all candidate translations to form another high potential area, indicating that there is a problem with the use of limit words, and so on. ​​​​​

[0076] Step S13, calling the audit engine to audit the initial text, and in the case where the audit result is that the initial text meets the preset condition, taking the initial text as the target text;

[0077] In this embodiment, after obtaining the initial text, further auditing and adjusting the text are needed to improve the translation quality.

[0078] Specifically, it includes:

[0079] Step S131, calling the audit engine to determine the flaw type and flaw level of the aggregated risk semantic cluster;

[0080] Step S132, according to the flaw type and the flaw level, performing a preset purification strategy on the text content corresponding to the aggregated risk semantic cluster, wherein the purification strategy includes uniform inconsistent term translation, replacing non-compliant text content, or intercepting unsafe text content;

[0081] Step S133, integrating the text content that meets the preset condition after the purification strategy processing with the text content of the non-flaw part in the candidate translation set, and generating the target text.

[0082] In this embodiment, first, the system analyzes each aggregated risk semantic cluster to determine its flaw type and level, and then processes it according to the preset purification strategy.

[0083] For example, for (inconsistent terms), the system determines its type to be "noun usage" and its level to be low after analysis, and then executes the "forced uniform" strategy. According to the knowledge base, it is determined that the official translation of "A project" is "Project A", so it sets the confidence of "Project A" to 1 and the confidence of "The a Project" to 0 in the final translation candidate pool, thereby locking the correct term; similarly for (compliance risk), the system determines its type to be "extreme word usage" and its level to be "high" after analysis, and according to the compliance strategy library, the library stipulates that for words such as "best", it should be replaced with "leading", so the system replaces the candidate translation "best" with "leading". In the text integration stage, the system processes other safe parts of the sentence and splices the above adjustment results to obtain the target text. Similarly.

[0084] Step S14, according to the target text, performing interface adaptive matching to display the target text;

[0085] In the embodiment, after the target language text is obtained, the text needs to be displayed on the user interface (UI) of the embedded device; since the length of the translation output is dynamically changed, static, fixed-size display areas are prone to have problems of truncated content or unattractive layout; in this regard, interface adaptive matching is performed according to the text content, thereby forming a complete closed-loop experience from content generation to human-computer interaction.

[0086] Specifically, the interface adaptive matching according to the target text comprises:

[0087] In step S141, length information of the target text is obtained.

[0088] In step S142, based on the length information, the size or layout of a display area in the user interface for displaying the target text is dynamically adjusted.

[0089] In the embodiment, after the target text is obtained, the interface adaptive module obtains the length information of the target language text, and based on the length information, the module dynamically adjusts the size or layout of a display area in the UI for displaying the text; for example, if the length of the text exceeds the default display area, the height of the area is automatically expanded, and the positions of adjacent UI elements are adjusted to ensure complete presentation of the content.

[0090] For example, the final English translation has a total of 180 characters, at this time, the interface adaptive module calculates that a height of 100 pixels is needed on the target screen to completely display, and the default height of the UI control for displaying the text is 40 pixels. At this time, the module sends an instruction to the UI framework to dynamically adjust the height of the control to 100 pixels, thereby ensuring that the user sees complete and clear translation results, and so on.

[0091] From the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software and a necessary general hardware platform, and of course, it can also be implemented by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk), and includes a plurality of instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device) to execute the method described in each embodiment of the present application.

[0092] An embedded translation system with a content review mechanism is also provided in the present embodiment, which is used to implement the above embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and contemplated.

[0093] Figure 2 FIG. 1 is a structural block diagram of an embedded translation system with a content review mechanism according to an embodiment of the present application, as shown in the figure, the system comprises: Figure 2

[0094] A request collection module 21 is configured to obtain a translation request, wherein the translation request is used to indicate a request to translate a text to be translated in a first language into a target text in a second language;

[0095] A translation module 22 is configured to call a translation interface to translate the text to be translated to obtain an initial text;

[0096] A text review module 23 is configured to call a review engine to review the initial text, and in a case where the review result is that the initial text meets a preset condition, the initial text is taken as a target text;

[0097] An interface matching module 24 is configured to perform interface adaptive matching according to the target text to display the target text.

[0098] In an optional embodiment, the calling of the translation interface to translate the text to be translated to obtain an initial text comprises:

[0099] Obtaining a text to be translated, and constructing a structured semantic representation based on the text to be translated;

[0100] Performing semantic mapping processing on the structured semantic representation to generate a candidate translation set containing potential semantic flaw units;

[0101] Performing risk aggregation evolution processing on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set;

[0102] Determining the initial text based on the aggregated risk semantic clusters.

[0103] In an optional embodiment, the performing of the risk aggregation evolution processing on the candidate translation set comprises:

[0104] Based on a preset knowledge base, calculating an initial risk potential value of each semantic flaw unit in the candidate translation set; ​

[0105] According to a preset iterative diffusion rule, the risk potential value is iteratively updated to obtain a target risk potential value;

[0106] The semantic flaw units with the target risk potential value exceeding a preset threshold are clustered to obtain the aggregated risk semantic cluster.

[0107] It should be noted that the above various modules can be realized by software or hardware, and for the latter, the following implementation manners can be used, but are not limited thereto: the above modules are located in the same processor; or the above various modules are located in different processors in any combination.

[0108] Embodiments of the present application also provide a computer readable storage medium, which stores a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0109] In an example embodiment, the above computer readable storage medium can include, but is not limited to, a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0110] Embodiments of the present application also provide an electronic device, which comprises a memory storing a computer program and a processor configured to execute the computer program to perform the steps in any of the above method embodiments.

[0111] In an example embodiment, the above electronic device can further comprise a transmission device connected to the processor and an input / output device connected to the processor.

[0112] Through the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional modules is taken as an example for illustration, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions.

[0113] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. For example, the division of the apparatus embodiments is merely an example, and for example, the division of the modules or units can be different, and for example, multiple modules or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0114] The units described as separate components may or may not be physically separate, and the components shown as units may be one physical unit or multiple physical units, i.e., may be located in one place, or may be distributed in multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0115] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0116] If the integrated unit is realized 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 solutions of the embodiments of the present application essentially or the parts that make contributions to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, and includes a number of instructions to make a device (which can be a single chip, a chip, etc.) or a processor execute all or part of the steps of the various embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0117] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any changes or replacements within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An embedded translation method with a content moderation mechanism, characterized in that, include: Obtain a translation request, wherein the translation request is used to instruct that the text to be translated in the first language be translated into the target text in the second language; The translation interface is invoked to translate the text to be translated, so as to obtain the initial text; The initial text is reviewed by the review engine, and if the review result indicates that the initial text meets the preset conditions, the initial text is used as the target text. The interface is adaptively matched based on the target text in order to display the target text. The process of calling the translation interface to translate the text to be translated to obtain the initial text includes: Obtain the text to be translated, and construct a structured semantic representation based on the text to be translated; Semantic mapping processing is performed on the structured semantic representation to generate a set of candidate translations containing potentially semantically flawed units; Risk clustering evolution processing is performed on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set; The initial text is determined based on the aggregated risk semantic cluster; The step of performing risk clustering and evolutionary processing on the candidate translation set includes: Based on a pre-defined knowledge base, the initial risk potential value of each semantically flawed unit in the candidate translation set is calculated; According to the preset iterative diffusion rule, the risk potential energy value is iteratively updated to obtain the target risk potential energy value; The semantic defective units whose target risk potential value exceeds a preset threshold are clustered to obtain the aggregated risk semantic cluster; Among them, in the first In the next iteration, the unit Risk potential value The update formula is: in, It is the simulation time step, which controls the update rate; yes The set of neighbors; These are weighting coefficients, representing the unit. right The intensity of the impact; The step of calling the review engine to review the initial text includes: The audit engine is invoked to determine the defect type and defect level of the aggregated risk semantic cluster; Based on the defect type and defect level, a preset purification strategy is executed on the text content corresponding to the aggregated risk semantic cluster. The purification strategy includes unifying inconsistent terminology translations, replacing non-compliant text content, or blocking unsafe text content. The text content that meets the preset conditions after being processed by the purification strategy is integrated with the non-flawed text content in the candidate translation set to generate the target text.

2. The method according to claim 1, characterized in that, The step of performing adaptive interface matching based on the target text includes: Obtain the length information of the target text; Based on the length information, the size or layout of the display area in the user interface used to display the target text is dynamically adjusted.

3. An embedded translation system with a content moderation mechanism, characterized in that, include: The request acquisition module is used to acquire translation requests, wherein the translation requests are used to instruct that the text to be translated in the first language be translated into the target text in the second language; The translation module is used to call the translation interface to translate the text to be translated in order to obtain the initial text; The text review module is used to call the review engine to review the initial text, and if the review result shows that the initial text meets the preset conditions, the initial text is used as the target text. The interface matching module is used to perform adaptive interface matching based on the target text in order to display the target text. The process of calling the translation interface to translate the text to be translated to obtain the initial text includes: Obtain the text to be translated, and construct a structured semantic representation based on the text to be translated; Semantic mapping processing is performed on the structured semantic representation to generate a set of candidate translations containing potentially semantically flawed units; Risk clustering evolution processing is performed on the candidate translation set to identify and locate one or more aggregated risk semantic clusters in the candidate translation set; The initial text is determined based on the aggregated risk semantic cluster; The step of performing risk clustering and evolutionary processing on the candidate translation set includes: Based on a pre-defined knowledge base, the initial risk potential value of each semantically flawed unit in the candidate translation set is calculated; According to the preset iterative diffusion rule, the risk potential energy value is iteratively updated to obtain the target risk potential energy value; The semantic defective units whose target risk potential value exceeds a preset threshold are clustered to obtain the aggregated risk semantic cluster; Among them, in the first In the next iteration, the unit Risk potential value The update formula is: in, It is the simulation time step, which controls the update rate; yes The set of neighbors; These are weighting coefficients, representing the unit. right The intensity of the impact; The step of calling the review engine to review the initial text includes: The audit engine is invoked to determine the defect type and defect level of the aggregated risk semantic cluster; Based on the defect type and defect level, a preset purification strategy is executed on the text content corresponding to the aggregated risk semantic cluster. The purification strategy includes unifying inconsistent terminology translations, replacing non-compliant text content, or blocking unsafe text content. The text content that meets the preset conditions after being processed by the purification strategy is integrated with the non-flawed text content in the candidate translation set to generate the target text.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to perform the method described in any one of claims 1 to 2 when executed.

5. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 2.

Images