AI large language model-based aided translation teaching system and method

Abstract

The invention belongs to the field of teaching auxiliary systems, and provides an auxiliary translation teaching system and method based on an AI large language model, and the method comprises the steps: outputting a first statement to a user, the language type of the first statement being a native language of the user; receiving a second statement of the user, wherein the second statement is the translation of the user on the first statement; generating a first prompt word according to the first statement and the second statement; receiving a reply of the A I large language model; analyzing the reply of the A I large language model according to the preset format; determining the correctness of user translation according to the analysis result; when the translation is incorrect, generating a second cue word; receiving a reply of the A I large language model, and analyzing the correct answer and the third exercise statement; and displaying the correct answer and the third exercise statement to the user, and prompting the user to translate the third exercise statement.

Description

Technical Field

[0001] The present invention belongs to the field of teaching assistance systems, and specifically relates to an auxiliary translation teaching system and method based on an AI large language model. Background Art

[0002] AI large language models are important AI models developed in recent years based on deep learning technology, specifically designed for processing and generating natural language. Pre-trained on large-scale corpora, these models possess billions or even hundreds of billions of parameters, enabling them to understand the complex semantics of language and generate contextually coherent textual expressions. Typical AI large language models include GPT Wenxin Yiyan, which have demonstrated outstanding performance across a variety of natural language processing tasks.

[0003] AI large language models are typically accessible to users through open developer APIs. Developers can obtain generated text content by passing input prompts and customize the length, style, and language characteristics of the generated text as needed. This capability makes AI large language models a core technology for a variety of application scenarios, such as dialogue systems, text generation, and translation tools.

[0004] Although AI large language models have performed well in the field of natural language processing, they still have some shortcomings in actual user use, especially in translation practice scenarios:

[0005] The output of AI large language models is highly dependent on the input prompts. During use, users often need to repeatedly adjust the content and format of the prompts to ensure that the model accurately understands the task requirements and generates the expected results. For example, in translation exercises, prompts need to clearly instruct the model to perform tasks such as correctness determination, error analysis, and answer generation. Otherwise, the model may output incomplete, inaccurate, or irrelevant content.

[0006] This process of adjusting prompt words has a high technical threshold for ordinary users, especially when users are not familiar with the working mechanism and language expression of the model. Adjusting prompt words is not only time-consuming but may also reduce usage efficiency. Summary of the Invention

[0007] In order to solve the problems in the prior art, the present invention provides an assisted translation teaching method based on an AI large language model, comprising the following steps:

[0008] Outputting a first sentence to the user, where the language type of the first sentence is the user's native language;

[0009] receiving a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user;

[0010] Generate a first prompt word based on the first sentence and the second sentence, wherein the first prompt word is used to inquire the AI large language model about the correctness of the translation of the second sentence and require the AI large language model to respond in a preset format;

[0011] Receiving a response from the AI large language model; parsing the response from the AI large language model according to the preset format; determining the correctness of the user's translation based on the parsing result; and when the translation is incorrect, generating a second prompt word, the second prompt word being used to request the AI large language model to generate a correct answer, generating a third practice sentence based on the error type, and replying in the second preset format;

[0012] Receive a response from the AI large language model, parse the correct answer and the third practice sentence; display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

[0013] Furthermore, the first sentence is entered through text input, voice input or through a handwriting tablet.

[0014] Furthermore, the first prompt word clearly specifies the output format of the AI large language model, including:

[0015] Determination of translation correctness;

[0016] If incorrect, specifically point out the error in the translation;

[0017] Provide correct translations;

[0018] Give brief correction suggestions or explanations.

[0019] Furthermore, the second prompt word includes the following information:

[0020] Original first statement;

[0021] User translations;

[0022] Specific types of translation errors and detailed descriptions;

[0023] The difficulty and context of the third practice sentence generated by AI are required.

[0024] Furthermore, the response content generated by the AI large language model based on the second prompt word is received, and the response is ensured to include the following information:

[0025] The third practice sentence;

[0026] The correct answer to the third exercise sentence;

[0027] Explanation of error types and corrections for practice sentences.

[0028] The present invention also provides an auxiliary translation teaching system based on an AI large language model, comprising the following modules:

[0029] an output module, configured to output a first sentence to a user, wherein the language type of the first sentence is the user's native language;

[0030] a receiving module, configured to receive a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user;

[0031] A first processing module is configured to generate a first prompt word based on the first sentence and the second sentence, wherein the first prompt word is used to inquire about the correctness of the translation of the second sentence by the AI large language model and to request the AI large language model to respond in a preset format;

[0032] a second processing module configured to receive a response from the AI large language model; parse the response from the AI large language model according to the preset format; determine the correctness of the user's translation based on the parsed result; and, if the translation is incorrect, generate a second prompt word, the second prompt word being used to request the AI large language model to generate a correct answer, generate a third practice sentence based on the error type, and respond in the second preset format;

[0033] The display module is used to receive the response of the AI large language model, parse the correct answer and the third practice sentence; display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

[0034] Furthermore, the first sentence is entered through text input, voice input or through a handwriting tablet.

[0035] Furthermore, the first prompt word clearly specifies the output format of the AI large language model, including:

[0036] Determination of translation correctness;

[0037] If incorrect, specifically point out the error in the translation;

[0038] Provide correct translations;

[0039] Give brief correction suggestions or explanations.

[0040] Furthermore, the second prompt word includes the following information:

[0041] Original first statement;

[0042] User translations;

[0043] Specific types of translation errors and detailed descriptions;

[0044] The difficulty and context of the third practice sentence generated by AI are required.

[0045] Furthermore, the response content generated by the AI large language model based on the second prompt word is received, and the response is ensured to include the following information:

[0046] The third practice sentence;

[0047] The correct answer to the third exercise sentence;

[0048] Explanation of error types and corrections for practice sentences.

[0049] This invention provides an assisted translation teaching method based on an AI large language model. By integrating advanced natural language processing technology and translation teaching needs, it overcomes many shortcomings of traditional translation teaching and has the following beneficial effects:

[0050] This invention utilizes an intelligent prompt word generation mechanism to dynamically generate appropriate prompt words based on user input and task requirements. This eliminates the complex process of frequently adjusting prompt words when using large AI language models, significantly lowering the barrier to entry and making it easy for even non-expert users to get started. The system presents translation results, error analysis, and targeted exercises through an intuitive interface and clear guidance, allowing users to complete translation learning tasks without having to master complex operational procedures.

[0051] The present invention is based on the secondary development of mature AI large language models (such as GPT, etc.), eliminating the need to build complex models from scratch, significantly reducing development time and resource costs. By combining the strong generalization capabilities of AI large language models, this system only requires a small amount of customized data to achieve high-quality translation feedback and practice generation, reducing data collection and annotation costs. The system architecture supports direct calls to the API interface of the large language model without the need for additional hardware deployment, making it easy to integrate into existing translation teaching platforms, thereby reducing development and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0054] Below, the invention is preferably described with reference to the accompanying drawings and specific embodiments.

[0055] An AI Large Language Model (ALM) is a large-scale AI model based on deep learning technology, specifically designed for processing and generating natural language. Models such as GPT and Wenxinyiyan are typically composed of billions or even hundreds of billions of parameters. Pre-trained on large-scale corpora, they can understand and generate semantically rich, contextually coherent language expressions.

[0056] AI large language models typically provide convenient access to users through developer APIs to support a variety of natural language processing tasks. Developers can obtain generated text content by passing input prompts, and can customize the length, style, and language of the generated text.

[0057] In one embodiment, reference Figure 1 Based on the existing AI large language model, the present invention provides an assisted translation teaching method based on the AI large language model. This method utilizes current advanced artificial intelligence technologies and the natural language processing and generation capabilities of large-scale pre-trained language models to address the core needs of translation teaching. This method incorporates a systematic approach that includes translation training, error feedback, language optimization, and personalized learning guidance. Through real-time interaction and intelligent analysis mechanisms, this method can effectively help users improve their translation skills, correct translation errors, and dynamically adjust teaching content based on the user's learning progress, thereby achieving an efficient and targeted translation teaching process.

[0058] Specifically, the method comprises the following steps:

[0059] Step S1: output a first sentence to a user, where the language type of the first sentence is the user's native language.

[0060] Based on the user's learning needs, source language sentences suitable for the current teaching objectives are randomly or purposefully selected from a predefined corpus. The topics of the sentences can cover different scenarios such as daily communication, professional fields, and cultural expressions, ensuring that the content is relevant to the user's learning goals;

[0061] The language difficulty of the first sentence is dynamically adjusted according to the user's current learning level, including but not limited to vocabulary complexity, sentence structure complexity and context relevance, to ensure that the sentence is both challenging and does not exceed the user's understanding.

[0062] The first sentence can be output to the user in a variety of forms, including text display, voice playback, or multimodal display combined with graphics and text, to enhance the user's learning experience and immersion.

[0063] When outputting the first sentence, ensure that the language type of the sentence is consistent with the user's native language, and at the same time combine the cultural background and expression habits of the native language so that the user can more easily understand the context and meaning of the sentence.

[0064] If the user has specific learning needs (such as professional translation in a certain field), the content theme and expression of the first sentence can be customized according to the user's needs, such as selecting scientific and technological sentences with more technical terms, or legal texts containing legal terms.

[0065] Step S2: receiving a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user.

[0066] The system provides multiple input methods for users to submit translation content, including text input, voice input, or input via a handwriting tablet;

[0067] When receiving voice or handwriting input, the system calls the corresponding voice recognition module or character recognition module to convert the input content into a standardized text format as the user's translation sentence.

[0068] Furthermore, after receiving the second sentence input by the user, the system first performs a language type detection on it to ensure that the input language meets the definition of the user's learning target language.

[0069] If it is detected that the language type of the input sentence does not match the target learning language (for example, the user mistakenly uses his native language or other languages), the system will immediately prompt and guide the user to re-enter.

[0070] Furthermore, the second sentence submitted by the user is preprocessed, including removing redundant characters (such as extra spaces and punctuation marks) and processing common input errors (such as spelling errors and irregular formats); in the case of voice or handwriting input, the context and the predictive ability of the language model are combined to automatically supplement or correct obvious input errors to improve the accuracy of sentence processing.

[0071] Furthermore, if the user is not satisfied with the translation content, the system allows the user to make multiple revisions and records the translation content of each revision for subsequent analysis of the user's translation ideas and progress.

[0072] Furthermore, when receiving user input, the system adapts the translation environment according to the learning scenario, such as quickly processing the user's translation content in real-time dialogue mode, or providing the user with sufficient modification time in non-real-time mode.

[0073] Furthermore, after receiving the user's second sentence, the system displays the translation content entered by the user through the interactive interface for the user to confirm; if the user confirms that it is correct, the system will officially record the translation sentence and enter the next processing step; if the user finds an error, he or she can choose to re-enter or use the intelligent suggestion function provided by the system to make adjustments.

[0074] Step S3: Generate a first prompt word based on the first sentence and the second sentence. The first prompt word is used to inquire the AI large language model about the correctness of the translation of the second sentence and require the AI large language model to reply in a preset format.

[0075] The first prompt word should clearly express to the AI large language model the need to determine whether the second sentence (target language translation) input by the user is consistent with the meaning of the first sentence (source language content);

[0076] The prompt word includes the specific language pair (such as the source language is Chinese and the target language is English) and contextual information for translation checking to ensure that the model accurately understands the task requirements.

[0077] The prompt words clearly define the output format of the AI large language model, including:

[0078] a) Determination of translation correctness (e.g., “correct” or “incorrect”);

[0079] b) If incorrect, please point out the specific errors in the translation (such as vocabulary usage, grammatical structure, deviation in sentence meaning, etc.);

[0080] c) provide a correct translation;

[0081] d) Provide brief correction suggestions or instructions to help users understand the cause of the error.

[0082] Furthermore, the prompt words can further include the user's learning background or current learning goals (such as focusing on practicing past tense, translating professional terms, etc.), so that AI can generate feedback that is more in line with user needs.

[0083] For example:

[0084] First sentence (source language): "He went to the library yesterday."

[0085] Second sentence (user translation, target language): "He went to the library yesterday." Generated first prompt:

[0086]

[0087] When the AI language model receives the above prompt, the content returned in the preset format may be as follows:

[0088]

[0089]

[0090] Step S4: receiving a response from the AI large language model; parsing the response from the AI large language model according to the preset format; determining the correctness of the user's translation based on the parsing result; and generating a second prompt word when the translation is incorrect, the second prompt word being used to require the AI large language model to generate a third practice sentence based on the error type and respond in a second preset format.

[0091] In this step, the reply content generated by the AI large language model based on the first prompt word is received. The reply contains information such as translation correctness judgment, error type analysis, correct answer and explanation in a preset format; the system ensures that the received reply is complete and meets the preset format requirements. If the reply format does not match, a feedback prompt will be generated and the AI reply will be requested again.

[0092] Parse the AI's response according to the preset format, extract key content (such as correctness judgment, error type, correction suggestions, etc.) as structured data for subsequent processing; classify the error types, including grammatical errors, vocabulary usage errors, sentence structure errors and semantic deviations, and record detailed error information.

[0093] The system determines whether the user's translation is correct based on the analysis results:

[0094] a) If the translation is correct, the user's translation result will be recorded and confirmed to the user;

[0095] b) If the translation is incorrect, a second prompt word is generated to further guide the AI to generate targeted practice content.

[0096] Based on the error type analyzed, a second prompt word is generated, prompting the AI to generate a third practice sentence based on the user's specific error;

[0097] The second prompt word includes the following information:

[0098] a) the original first sentence;

[0099] b) User translation (second sentence);

[0100] c) Specific types and detailed descriptions of translation errors;

[0101] d) The difficulty and context of the third practice sentence generated by AI.

[0102] The second prompt word clarifies the output requirements of the AI, including that the generated third practice sentence must be targeted, the error explanation must be related to the practice sentence, and example answers or explanations must be provided.

[0103] Exemplarily, the example in step S3 is continued.

[0104] The system receives the above AI response and parses it to extract the following structured content:

[0105] Translation accuracy: Incorrect.

[0106] Error type: tense error (wrong verb form).

[0107] Correct answer: He went to the library yesterday.

[0108] Error explanation: The verb "go" needs to be changed to the past tense "went" to match the time marker "yesterday".

[0109] Based on the error type analyzed, the system generates a second prompt, requiring the AI to design exercises targeting tense errors:

[0110] {

[0111] "input":"The user made a translation error.

[0112] The source sentence is:'He went to the library yesterday.'

[0113] The user's translation was:'He go to the library yesterday.'

[0114] The error is related to tense.

[0115] Please generate a third practice sentence that helps the user focus on practicing past tense verbs. The practice sentence should be related to the original context and include the correct answer and an explanation."

[0116] }

[0117] AI generates the following exercises based on the prompts:

[0118]

[0119] Step S5: Receive a response from the AI large language model, parse the correct answer and the third practice sentence, display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

[0120] In this step, the response content generated by the AI large language model based on the second prompt word is received, ensuring that the response contains the following information:

[0121] a) The third practice sentence;

[0122] b) The correct answer to the sentence in the third exercise;

[0123] c) Explanation of error types and corrections for practice sentences.

[0124] Verify the completeness of the response format. If any critical information is missing, send error feedback to the AI and re-request.

[0125] The system parses the AI response, extracts the third practice sentence and its related information, and stores them in categories:

[0126] a) The third practice sentence: for user translation;

[0127] b) Correct answer: For user reference;

[0128] c) Correction explanation: for users to understand the cause of the error and improve their translation skills.

[0129] Based on the parsing results, the correct answer and the third exercise sentence are displayed in a user-friendly manner;

[0130] For practice sentences, the system provides guiding prompts (such as paying attention to using the correct verb tense, sentence structure, etc.) to help users focus on their current weaknesses.

[0131] The following information is displayed to the user:

[0132] a) The correct translation result of the first sentence;

[0133] b) the content of the third practice sentence;

[0134] c) Explicitly ask the user to translate the third practice sentence and submit it after completion.

[0135] For example:

[0136] Receiving and parsing AI responses

[0137] Parse the following:

[0138] Correct answer (translation of the first sentence):

[0139] "He went to the library yesterday."

[0140] The third practice sentence:

[0141] "She walked to the park last weekend."

[0142] Exercise explanation:

[0143] This sentence practice aims to strengthen the use of past tense verbs. 'walked' is the past tense form of 'walk,' which meets the tense requirements of the time adverbial 'last weekend.'

[0144] Content displayed to users

[0145] The correct translation of the previous sentence is:

[0146] "He went to the library yesterday."

[0147] Please translate the following sentences:

[0148] "She walked to the park last weekend."

[0149] hint:

[0150] Pay attention to practicing the use of past tense verbs, such as the use of 'went' in the first sentence.

[0151] On the other hand, the present invention also provides an auxiliary translation teaching system based on an AI large language model, comprising:

[0152] an output module, configured to output a first sentence to a user, wherein the language type of the first sentence is the user's native language;

[0153] a receiving module, configured to receive a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user;

[0154] A first processing module is configured to generate a first prompt word based on the first sentence and the second sentence, wherein the first prompt word is used to inquire about the correctness of the translation of the second sentence by the AI large language model and to request the AI large language model to respond in a preset format;

[0155] a second processing module configured to receive a response from the AI large language model; parse the response from the AI large language model according to the preset format; determine the correctness of the user's translation based on the parsed result; and, if the translation is incorrect, generate a second prompt word, the second prompt word being used to request the AI large language model to generate a correct answer, generate a third practice sentence based on the error type, and respond in the second preset format;

[0156] The display module is used to receive the response of the AI large language model, parse the correct answer and the third practice sentence; display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

[0157] The prior art mentioned in the above background technology section and specific embodiments section of the present invention can be regarded as part of the present invention and used to understand the meaning of some technical features or parameters.

Claims

1. An assisted translation teaching method based on an AI large language model, characterized in that: The method comprises the following steps: Outputting a first sentence to the user, where the language type of the first sentence is the user's native language; receiving a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user; Generate a first prompt word based on the first sentence and the second sentence, wherein the first prompt word is used to inquire the AI large language model about the correctness of the translation of the second sentence and require the AI large language model to respond in a preset format; Receiving a response from the AI large language model; parsing the response from the AI large language model according to the preset format; determining the correctness of the user's translation based on the parsing result; and when the translation is incorrect, generating a second prompt word, the second prompt word being used to request the AI large language model to generate a correct answer, generating a third practice sentence based on the error type, and replying in the second preset format; Receive a response from the AI large language model, parse the correct answer and the third practice sentence; display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

2. The AI large language model-based assisted translation teaching method according to claim 1 is characterized in that: The first sentence is entered through text input, voice input or through a handwriting tablet.

3. The AI large language model-based assisted translation teaching method according to claim 1 is characterized in that: The first prompt word clearly specifies the output format of the AI large language model, including: Determination of translation correctness; If incorrect, specifically point out the error in the translation; Provide correct translations; Give brief correction suggestions or explanations.

4. The AI large language model-based assisted translation teaching method according to claim 1 is characterized in that: The second prompt word includes the following information: Original first statement; User translations; Specific types of translation errors and detailed descriptions; The difficulty and context of the third practice sentence generated by AI are required.

5. The AI large language model-based assisted translation teaching method according to claim 1 is characterized in that: Receive the response generated by the AI large language model based on the second prompt word, ensuring that the response contains the following information: The third practice sentence; The correct answer to the third exercise sentence; Explanation of error types and corrections for practice sentences.

6. An assisted translation teaching system based on an AI large language model, characterized in that: The system includes the following modules: an output module, configured to output a first sentence to a user, wherein the language type of the first sentence is the user's native language; a receiving module, configured to receive a second sentence of the user, where the second sentence is a translation of the first sentence by the user, and the language of the second sentence is the target learning language of the user; A first processing module is configured to generate a first prompt word based on the first sentence and the second sentence, wherein the first prompt word is used to inquire about the correctness of the translation of the second sentence by the AI large language model and to request the AI large language model to respond in a preset format; a second processing module configured to receive a response from the AI large language model; parse the response from the AI large language model according to the preset format; determine the correctness of the user's translation based on the parsed result; and, if the translation is incorrect, generate a second prompt word, the second prompt word being used to request the AI large language model to generate a correct answer, generate a third practice sentence based on the error type, and respond in the second preset format; The display module is used to receive the response of the AI large language model, parse the correct answer and the third practice sentence; display the correct answer and the third practice sentence to the user, and prompt the user to translate the third practice sentence.

7. The AI large language model-based auxiliary translation teaching system according to claim 6 is characterized in that: The first sentence is entered through text input, voice input or through a handwriting tablet.

8. The AI large language model-based auxiliary translation teaching system according to claim 6 is characterized in that: The first prompt word clearly specifies the output format of the AI large language model, including: Determination of translation correctness; If incorrect, specifically point out the error in the translation; Provide correct translations; Give brief correction suggestions or explanations.

9. The AI large language model-based auxiliary translation teaching system according to claim 6 is characterized in that: The second prompt word includes the following information: Original first statement; User translations; Specific types of translation errors and detailed descriptions; The difficulty and context of the third practice sentence generated by AI are required.

10. The AI large language model-based auxiliary translation teaching system according to claim 6 is characterized in that: Receive the response generated by the AI large language model based on the second prompt word, ensuring that the response contains the following information: The third practice sentence; The correct answer to the third exercise sentence; Explanation of error types and corrections for practice sentences.

Images