Traditional Chinese medicine cross-language aided translation system based on artificial intelligence

Abstract

The invention provides a traditional Chinese medicine cross-language aided translation system based on artificial intelligence, and belongs to the technical field of traditional Chinese medicine translation. The traditional Chinese medicine cross-language aided translation system based on artificial intelligence comprises a data acquisition module, a semantic recognition module, a semantic correction module, a translation strategy output module and a translation evaluation module. The system can track semantic distance changes between term translation and historical standards in real time, a translation migration file is established for recognized deviations, the corresponding relation between word meaning drift paths and contexts is recorded, the system can judge whether the deviations belong to common error types or not, rapid error correction is conducted on the basis of a term evolution database, and therefore the accuracy of term translation is improved. The mechanism does not depend on a fixed dictionary rule, but performs combined judgment according to factors such as a semantic change trend, a context structure adaptation degree and the frequency of historical deviation behaviors, so that the intelligence and the adaptability of deviation judgment are greatly improved, and the self-correction capability and the fault tolerance of a translation system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of TCM translation technology, and more specifically, to an artificial intelligence-based cross-language assisted translation system for TCM. Background Technology

[0002] Traditional Chinese medicine (TCM), a treasure of Chinese traditional culture, has a history of thousands of years. Based on the philosophical concept of "harmony between man and nature," it emphasizes the theories of Yin and Yang and the Five Elements, the theory of Zang-Fu organs and meridians, and the principle of diagnosis and treatment based on syndrome differentiation. The TCM system not only includes a rich theoretical framework but also encompasses multiple fields such as pharmacology, formulary, acupuncture, and massage. Chinese herbal medicines, as an important component of TCM, possess unique characteristics in their pharmacological theories, mechanisms of action, and dosage methods, and are widely used globally. With the inheritance and development of TCM, more and more Chinese herbal medicines are being used in clinical treatment and are gradually going international. The internationalization of TCM is accelerating, especially in treating chronic diseases, improving quality of life, and preventing diseases, demonstrating unique advantages. However, with the international development of TCM, translation issues have gradually become a pressing problem. The translation of TCM names must not only accurately express the efficacy of the drugs but also convey their cultural background, historical connotations, and regional characteristics. Cross-language translation of Chinese herbal medicine names is a core link in the global dissemination of TCM, and the challenges and complexities of this translation process have spurred the demand for intelligent translation technology.

[0003] Current translation systems for Chinese medicinal herbs (TCM) names primarily rely on traditional methods such as static dictionary translation and rule-based translation models. These systems typically employ dictionary lookups or pre-defined translation rules, enabling them to handle some standardized translation tasks to a certain extent. However, they often have significant limitations when faced with complex TCM names. First, TCM names may have different expressions in different cultural backgrounds and languages, and existing systems fail to fully consider regional differences, dialect differences, and the transmission of cultural connotations. Second, many TCM names are not merely simple drug names but also contain rich historical, cultural, and pharmacological information, which cannot be accurately conveyed through traditional translation systems, resulting in translations that may fail to reflect the deeper meaning of the original text. Furthermore, existing translation systems largely depend on static data and lack dynamic update mechanisms, making it difficult to adapt to the ever-evolving terminology and translation needs in the field of TCM. Summary of the Invention

[0004] To overcome the above deficiencies, the present invention provides an artificial intelligence-based cross-language assisted translation system for traditional Chinese medicine that overcomes or at least partially solves the above technical problems.

[0005] This invention is implemented as follows: This invention provides an artificial intelligence-based cross-language assisted translation system for traditional Chinese medicine, including a data acquisition module, a semantic recognition module, a semantic correction module, a translation strategy output module, and a translation evaluation module; The data acquisition module is used to collect multi-source data on the names of Chinese medicinal materials, including vocabulary attribute information, historical literature data, regional cultural background data and pronunciation features, and to perform multi-dimensional vector representation and normalization processing on the collected data. The semantic recognition module, based on natural language processing technology, constructs a semantic relationship network of Chinese medicinal herb names and extracts biological attribute factors from these names. Regional evolution factors and hierarchical semantic factors And establish a comprehensive semantic vector ; The semantic correction module is used to modify the comprehensive semantic vector. Combined with historical evolution error function Sensitivity coefficient for contextual ambiguity The correction factor is calculated, and the identification bias is corrected through the error accumulation function to obtain the corrected comprehensive semantic vector. ; The translation strategy output module is used to output the translation strategy based on the corrected comprehensive semantic vector. By combining dictionary databases and standard translation databases, translations of the target language are generated according to preset rules. At the same time, it generates translations in the target language. Perform structural optimization and semantic matching adjustments; The translation evaluation module is used to evaluate the translation using a cosine similarity function. and credibility calculation function To generate target language translations The system performs a credibility score and outputs a classification of credibility levels, including high credibility, medium credibility, and low credibility, and provides corresponding correction suggestions.

[0006] In a preferred embodiment, the data acquisition module includes a lexical attribute extraction unit, a historical corpus integration unit, and a data preprocessing unit; The vocabulary attribute extraction unit obtains the pinyin, stroke count, part-of-speech tagging, and semantic labels of Chinese medicine names through natural language processing technology; The historical corpus integration unit integrates data from ancient books, modern documents, and folk terms for Chinese medicine names to form a time-series corpus database. The data preprocessing unit performs missing value filling, noise filtering, and multi-dimensional normalization on the collected multi-source data to ensure the accurate operation of the subsequent semantic recognition module.

[0007] In a preferred embodiment, the semantic recognition module includes a semantic inference unit and a semantic connection unit. The semantic inference unit employs a semantic relation inference model based on graph neural networks to establish semantic connection weights between words. And calculate the nodes using the formula. The semantic vector representation; The semantic joint unit uses biological attribute factors Regional evolution factors and hierarchical semantic factors Joint encoding to form a comprehensive semantic vector .

[0008] In a preferred embodiment, the semantic correction module includes a semantic correction unit, a translation accuracy detection unit, and an automatic translation result correction unit. The semantic correction unit is used to utilize a historical evolution error function. ; And combined with the contextual ambiguity sensitivity coefficient We calculate the sensitivity of semantic vectors to contextual changes and adjust the translation strategy accordingly.

[0009] In a preferred embodiment, the translation strategy output module employs a function mapping model to generate standard translation vectors that conform to semantic and cultural context. .

[0010] In a preferred embodiment, the translation evaluation module includes a similarity calculation unit, a state determination unit, and a response strategy unit. The similarity calculation unit is based on the cosine similarity calculation formula. Calculate the standard translation vector With the system output translation vector And combined with the credibility calculation function Calculate the standard translation vector With the system output translation vector Credibility.

[0011] In a preferred embodiment, the translation accuracy detection unit and the translation result automatic correction unit are included. The translation accuracy detection unit evaluates the translation results output by the recognition module from multiple dimensions based on three indicators: semantic similarity, term matching degree, and contextual consistency, and sets a first accuracy threshold, a second accuracy threshold, and a third accuracy threshold. When the semantic similarity is ≥0.90, the term matching degree is ≥0.85, and the contextual consistency is ≥0.88, the translation result is determined to be in the high accuracy safe range and no correction is required. When the semantic similarity is between 0.75 and 0.90, the term matching degree is between 0.65 and 0.85, and the contextual consistency is between 0.70 and 0.88, the translation result is determined to be in the warning range. The accuracy of the warning range is lower than that of the safe range, triggering the automatic correction unit to perform terminology standardization replacement and sentence optimization. When the semantic similarity is below 0.75, the term matching degree is below 0.65, or the contextual consistency is below 0.70, the translation result is determined to be in the danger zone. The accuracy of the danger zone is lower than that of the warning zone, and the system automatically starts multiple rounds of translation model iteration and manual verification prompts.

[0012] In a preferred embodiment, the automatic translation result correction unit includes a terminology standardization subunit, a grammatical structure adjustment subunit, and a human feedback fusion subunit; The terminology standardization subunit automatically replaces non-standard or easily confused professional terms in translation by calling a pre-set database of traditional Chinese medicine terminology, and ensures the rationality of the replacement through corpus verification. The grammar structure adjustment subunit uses natural language processing technology to optimize sentence structure and word order to ensure that the translation conforms to the expression habits of the target language. The human feedback fusion subunit is used to receive feedback from human experts on the correction results, and combines machine learning algorithms to dynamically update the translation model parameters and terminology database, forming a continuously iterative intelligent correction mechanism.

[0013] In a preferred embodiment, the state determination unit sets a safety threshold, a warning threshold, and a danger threshold based on the accuracy evaluation results output by the semantic correction module, and classifies the translation results into three categories: safe state, warning state, and critical state. The safety interval corresponds to an accuracy index ≥0.90, the translation results are stable, the system maintains the current translation model and terminology database, and no additional intervention is required; The warning range corresponds to an accuracy index between 0.75 and 0.90. The system will issue a warning, indicating that the translation result may have flaws and suggesting manual review or auxiliary correction. When the accuracy index in the danger zone falls below 0.75, the system issues a danger alarm, initiates multi-model integrated translation and expert manual review to prevent the spread of erroneous translations.

[0014] In a preferred embodiment, the response strategy unit is used to output different strategies for each threshold interval, including: When within the safe range, the system maintains the current translation algorithm and terminology database, and regularly updates and backtracks the quality of Chinese medicine professional corpus offline, without the need for immediate intervention; During the warning period, the system automatically triggers the semantic enhancement algorithm to perform secondary optimization translation based on the context, and automatically generates a manual review task to ensure that the review feedback cycle does not exceed 24 hours; In critical situations, the system immediately activates a multi-engine fusion translation strategy, combining rule matching and neural network deep learning technology to perform multiple rounds of correction, automatically label potential errors and high-risk terms, and generate a detailed correction suggestion report.

[0015] This invention provides an artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine, the beneficial effects of which include: 1. By constructing an intelligent Chinese-English translation system for traditional Chinese medicine materials that integrates data collection, identification, correction, and strategy output, the accuracy of terminology consistency recognition has been significantly improved. This fundamentally solves the problems of semantic ambiguity, word meaning deviation, and terminology ambiguity in cross-language communication in the field of traditional Chinese medicine. Due to their long history, frequent semantic evolution, and large regional differences, traditional Chinese medicine terms often have multiple translations for one term or multiple terms sharing the same translation. This can easily lead to "inaccurate translation" or "semantic drift" under manual translation or general translation engines. However, this system adopts a technical approach of "historical evolution record + dynamic semantic vector recognition + contextual logic modeling," which can perform high-precision tracing and accurate matching of the names of traditional Chinese medicine materials at the semantic level.

[0016] 2. During the translation process, the system tracks the semantic distance between the translated terms and historical standards in real time. It establishes a "translation offset profile" for identified deviations, recording their semantic drift paths and contextual correspondences. Simultaneously, the system determines whether the deviation belongs to common error types, such as semantic swapping, improper use of terminology polysemous words, or spelling interference. Based on the terminology evolution database, it performs rapid error correction. This mechanism does not rely on fixed dictionary rules but combines factors such as semantic change trends, contextual structure fit, and the frequency of historical deviation behaviors to make judgments. This significantly improves the intelligence and adaptability of deviation judgment, and greatly enhances the self-correction ability and fault tolerance of the translation system, providing users with higher quality and more interpretable translation results.

[0017] 3. Through settings, the system will automatically classify the credibility of the translation after completion. Based on indicators such as semantic matching degree, contextual coherence, and translation deviation threshold, the translation results will be divided into high credibility zone, medium warning zone, and low credibility risk zone, and corresponding processing strategies will be output for each zone. When the system judges that the translated terminology does not match the context, or contains expressions that have been marked as deviations in the past, a medium warning prompt will be triggered, suggesting that the user conduct a manual review. If it is judged as a high-risk deviation, such as excessive semantic drift of key terms or serious contextual conflict, it will be automatically marked as a risk zone. The system will actively prevent the publication or delivery of the translation and call the alternative strategy model to regenerate a corrected version. It has significant technical advantages in ensuring the accurate transmission of TCM terminology and reducing the risk of cross-contextual misunderstanding caused by translation errors, demonstrating a high level of practicality, foresight, and intelligence. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is an overall flowchart provided by an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 The present invention provides a technical solution: a cross-language assisted translation system for traditional Chinese medicine based on artificial intelligence, including a data acquisition module, a semantic recognition module, a semantic correction module, a translation strategy output module and a translation evaluation module; The data acquisition module is used to collect multi-source data on the names of Chinese medicinal materials, including vocabulary attribute information, historical literature data, regional cultural background data and pronunciation features, and to perform multi-dimensional vector representation and normalization processing on the collected data. In this system, the data acquisition module is a fundamental and crucial component. This module is responsible for collecting multi-dimensional data related to the names of Chinese medicinal herbs from multiple sources, including but not limited to historical documents, traditional medical materials, modern pharmacological research findings, and cultural background information. Through comprehensive data acquisition of Chinese medicinal herb names, the system can provide rich raw data for subsequent semantic analysis.

[0022] The semantic recognition module, based on natural language processing technology, constructs a semantic relationship network of Chinese medicinal herb names and extracts biological attribute factors from these names. Regional evolution factors and hierarchical semantic factors And establish a comprehensive semantic vector ; The data collected by the data acquisition module is transmitted to the semantic recognition module. This module employs natural language processing techniques, particularly deep learning-based neural network models, to perform semantic analysis on the names of Chinese medicinal herbs. The core task of the semantic recognition module is to extract the inherent biological attribute factors, regional evolution factors, and cultural background factors from the names of Chinese medicinal herbs. To achieve this, the system maps the names of Chinese medicinal herbs into high-dimensional semantic vectors and uses deep learning models to perform semantic relationship reasoning to capture the meaning of each Chinese medicinal herb name.

[0023] The semantic correction module is used to modify the comprehensive semantic vector. Combined with historical evolution error function Sensitivity coefficient for contextual ambiguity The correction factor is calculated, and the identification bias is corrected through the error accumulation function to obtain the corrected comprehensive semantic vector. ; The initial translation results generated by the semantic recognition module may contain some errors or biases, especially considering the polysemous words, synonyms, and regional differences in traditional Chinese medicine culture. The semantic correction module automatically corrects the translation results by learning from historical data and combining it with contextual information. This module employs an adaptive error correction algorithm, using multiple rounds of feedback and correction to make the final translation result more accurate and precise.

[0024] The translation strategy output module is used to output the translation strategy based on the corrected comprehensive semantic vector. By combining dictionary databases and standard translation databases, translations of the target language are generated according to preset rules. At the same time, it generates translations in the target language. Perform structural optimization and semantic matching adjustments; The translation strategy output module is the core of the entire system. Based on corrected semantic vectors and historical corpus data, it translates the names of traditional Chinese medicines into the target language. This module goes beyond simple word substitution, taking into account the cultural background and context of the target language. To this end, the translation strategy output module uses advanced machine translation technology, employing multi-level semantic mapping to ensure that the translation results not only conform to grammatical rules but also convey the cultural connotations of traditional Chinese medicines.

[0025] The core advantage of the entire system lies in its high degree of automation and intelligence, which overcomes the limitations of traditional translation methods. For example, traditional dictionary translation methods often ignore cultural background and contextual differences, while artificial intelligence systems, through learning from large amounts of data, can effectively capture the precise meaning of Chinese medicine names in different contexts.

[0026] The synergy of these modules achieves the following beneficial effects: Traditional Chinese medicine name translation often relies on manual translation, which is prone to misunderstandings or inaccuracies due to cultural differences. By introducing artificial intelligence technology, the system can automatically analyze and correct the translation of Chinese medicine names, ensuring high accuracy. Especially when dealing with polysemous and synonymous words, the system can accurately grasp the meaning of the translation in different contexts through semantic reasoning.

[0027] Handling complex semantics and cultural context: The names of Chinese medicines are not merely simple words; they often contain profound cultural backgrounds and historical significance. Traditional translation methods cannot fully understand and convey this cultural connotation, while AI translation systems, trained on large-scale corpora, can better handle differences in cultural background and provide translations that are more in line with the target language and culture.

[0028] Reduced human intervention: Traditional translation work requires a significant amount of human involvement, especially when accurately translating the names of traditional Chinese medicines. Artificial intelligence systems can automatically handle large volumes of translation work, reducing human intervention and making the translation process more efficient. Furthermore, they can operate 24 / 7, greatly improving translation efficiency.

[0029] Intelligent learning and adaptive capabilities: Through machine learning and deep learning technologies, the artificial intelligence system can continuously optimize and improve translation quality. During the translation process, the system constantly adjusts its translation strategy based on feedback, gradually enhancing its adaptive capabilities. This is particularly evident when dealing with new or unfamiliar Chinese medicine names, where the system can make reasonable inferences based on existing knowledge.

[0030] Enhancing the convenience of cross-language communication: With the global spread of Traditional Chinese Medicine (TCM) culture, the demand for cross-language communication in TCM is increasing. This translation system can effectively support international communication in the field of TCM, especially in the translation of Chinese medicinal herb names, ensuring consistent understanding across different languages ​​and promoting the globalization of TCM.

[0031] Reduced Translation Costs: Traditional Chinese medicine translation typically requires specialized personnel and involves lengthy communication and coordination processes. This AI-powered translation system allows businesses and research institutions to save significant human and time costs, while also improving the standardization and consistency of translations.

[0032] Enhancing Cross-Cultural Communication: This system is not merely a translation tool, but a bridge for cross-cultural exchange. By fully understanding and conveying the essence of Traditional Chinese Medicine (TCM), the system can help people from different cultural backgrounds better understand the core principles of TCM, promoting cultural communication and cooperation globally.

[0033] Scalability and Flexibility: The system is designed with excellent scalability, allowing for customization to meet user needs. Users can adjust the system's translation strategies or data sources according to different language requirements and application scenarios, making it more adaptable to various TCM translation needs.

[0034] Multi-dimensional processing of Chinese medicine names: This system not only focuses on single language conversion, but also ensures the depth and comprehensiveness of the translation results by analyzing Chinese medicine names from multiple dimensions, such as biological attributes, regional background, historical culture and other information, thus avoiding the superficial translation that is limited to traditional translation methods.

[0035] The translation evaluation module is used to evaluate the translation using a cosine similarity function. and credibility calculation function To generate target language translations The system performs a credibility score and outputs a classification of credibility levels, including high credibility, medium credibility, and low credibility, and provides corresponding correction suggestions.

[0036] The data acquisition module includes a vocabulary attribute extraction unit, a historical corpus integration unit, and a data preprocessing unit. The vocabulary attribute extraction unit obtains the pinyin, stroke count, part-of-speech tagging, and semantic labels of Chinese medicine names through natural language processing technology; The historical corpus integration unit integrates data from ancient books, modern documents, and folk terms for Chinese medicine names to form a time-series corpus database. The data preprocessing unit performs missing value filling, noise filtering, and multi-dimensional normalization on the collected multi-source data to ensure the accurate operation of the subsequent semantic recognition module.

[0037] In the above, this module is completed collaboratively by three sub-modules: the lexical attribute extraction unit, the historical corpus integration unit, and the data preprocessing unit: The lexical attribute extraction unit mainly extracts diverse traditional Chinese medicine name information such as "Coptis chinensis" (noun, name of a medicinal herb) and "Fried Coptis chinensis" (verb-object structure) by means of word-pos tagging, named entity recognition (NER), pinyin analysis, etc.; the historical corpus integration unit uses a time-annotated corpus (such as from "Treatise on Febrile Diseases" to "Chinese Pharmacopoeia"), and introduces the integration of folk corpora (such as folk remedies and local medicine records) to construct a time-evolution corpus database. The data preprocessing unit is responsible for processing redundant data, and uses algorithms such as normalization, missing value filling, standardization, and noise reduction to improve the quality of the corpus and provide clean input for semantic recognition; Through the fusion of structured and unstructured data, improve the digital representation ability of traditional Chinese medicine knowledge; Adding a time dimension to corpus integration helps to solve the problem of "historical translation drift". For example, the word "Aquilariae Lignum" was translated as "aloes wood" in the Tang Dynasty, and is more inclined to "agarwood" in modern times; the data preprocessing mechanism improves the overall fault tolerance of the system, enabling the system to still ensure high recognition accuracy in the presence of dialect synonyms, spelling ambiguities, etc.; enhances the system's ability to process rare traditional Chinese medicine noun entries (such as different local species under "Pinelliae Rhizoma"); the constructed multi-source time-series database provides a solid foundation for subsequent semantic evolution analysis and correction.

[0038] The semantic recognition module includes a semantic inference unit and a semantic combination unit. The semantic inference unit uses a semantic relationship inference model based on a graph neural network to establish the semantic connection weights between words and calculates the semantic vector representation of node through the following formula: ; where, is the representation of the -th layer node , is the neighbor set of node , is the degree of node , is the activation function, usually taking ; The semantic combination unit jointly encodes through the biological attribute factor , the regional evolution factor and the hypernym and hyponym semantic factor to form a comprehensive semantic vector .

[0039] In the above, taking the word "Chuanxiong Rhizoma" as an example, the initial word vector is set as Its neighbor set nodes include "Xiongqiong" (similar in pronunciation) and "Angelica sinensis" (common compatibility), and their vectors are respectively , ; Then the comprehensive semantic vector is updated to , and the calculation is as follows: ; The semantic recognition module is one of the core parts of the entire Chinese medicine cross - language translation system. It is responsible for conducting in - depth semantic analysis and processing on the collected Chinese herbal medicine names to ensure the accuracy and precision in the translation process. Different from traditional translation systems that rely on vocabulary comparison, the work of the semantic recognition module is more complex. It needs to consider the diversity of languages, cultural backgrounds, and the multiple meanings of Chinese medicine names in order to make accurate translation decisions. This module will input the name of each Chinese herbal medicine into a pre - trained language model and analyze its context relationship through a deep neural network to understand the true meaning behind the name.

[0040] The semantic recognition module conducts semantic analysis on Chinese medicine names through deep learning technology, avoiding the simple word - for - word replacement in traditional translation methods and being able to better understand the complex meanings of Chinese medicine names. Through context analysis, the handling of polysemous words and synonyms, the semantic recognition module greatly improves the accuracy of translation results, making the translation not only conform to language norms but also more natural and fluent. The semantic recognition module can automatically handle these polysemous words and synonyms problems through technologies such as context semantic analysis and knowledge graphs, thus avoiding misunderstandings and misinterpretations that may occur in traditional translation methods. By integrating a large amount of cultural background information and historical materials, the semantic recognition module can accurately understand the cultural connotations of Chinese medicine names, avoiding cultural conflict problems that occur in traditional translations and making the translation results more in line with the needs of the target language and culture.

[0041] Generally speaking, as a core component of the Chinese medicine cross - language translation system, the semantic recognition module undertakes a very important task. Through various advanced technologies such as deep learning technology, context semantic analysis, and knowledge graph construction, the system can accurately understand and analyze the multiple semantics of Chinese medicine names and continuously improve the translation accuracy through optimized algorithms.

[0042] The semantic correction module includes a semantic correction unit, a translation accuracy detection unit, and a translation result automatic correction unit. The semantic correction unit is used to utilize the historical evolution error function , defined as: ; Among them, is the value of the currently predicted semantic vector in the dimension, is in a specific historical period The standard semantic vector value is the time node factor, which dynamically corrects the semantic recognition deviation through a function and is the dimension of the semantic vector; In the above, the semantic inference unit models the semantic network graph based on graph neural networks (such as GCN, GAT), and constructs a graph for the hyponymy, correlation, and synonymy relationships between words; the semantic combination unit jointly vectorizes the "biological attribute factor + regional evolution factor + hyponymy semantic factor", avoiding the limitations of single word embedding vector representation, accurately capturing the multi-dimensional context of the meaning of traditional Chinese medicine, and the semantic combination vector mechanism can dynamically adjust the semantic interpretation of the same term according to the context, effectively handling sentence pattern differences, and providing more accurate semantic basic data for the subsequent semantic correction and translation strategy module.

[0043] Taking "Licorice" in traditional Chinese medicine as an example, assume the predicted semantic vector is: The historical standard semantic vector is: The time node factor , then the calculation result is: ; It indicates that there is a moderate deviation between the meaning of "Licorice" predicted by the system and the historical meaning.

[0044] Combined with the context ambiguity sensitivity coefficient , through the following formula: ; is the target semantic embedding function, is the word vector representation of the context input sentence.

[0045] Calculate the sensitivity of the semantic vector to context changes and adjust the translation strategy.

[0046] In the above, taking "Coptis" in traditional Chinese medicine as an example, considering two context sentences, 1. "Coptis has the effect of clearing heat and drying dampness"; 2. "Coptis has the efficacy of purging fire and detoxifying", after the system encodes through the context ambiguity sensitivity coefficient , the generated semantic vectors are shown in the following table:

[0047] In the context of modern literature, the semantic change is stable (the derivative is small), indicating that the meaning of "Coptis" is stable; In the context of integrating ancient books, the derivative is still small, indicating that the meaning of "Coptis" is stable; Therefore, the system translates "Coptis" as Coptis chinensis.

[0048] The semantic correction module includes a semantic correction unit, a translation accuracy detection unit, and an automatic translation result correction unit. The semantic correction unit utilizes a historical evolution error function, combined with a time-node factor, to correct deviations in semantic recognition. The translation accuracy detection unit performs multi-dimensional evaluation of the translation results based on semantic similarity, terminology matching, and contextual consistency, and determines whether the translation results meet high accuracy standards based on set thresholds. If not, the translation results will be automatically corrected. Through historical evolution error correction and contextual models, the accuracy of the translation is significantly improved. Multi-dimensional accuracy detection provides more precise translation quality control, and the automatic correction function improves translation efficiency, reduces manual intervention, and adapts to the complexity of TCM translation.

[0049] The translation strategy output module adopts a function mapping model. : ; in, This is the corrected comprehensive semantic vector. For contextual constraints, As an authoritative dictionary knowledge base, functions Generate target language translations that conform to semantic and cultural context by combining multi-factor mapping rules. .

[0050] The translation evaluation module includes a similarity calculation unit, a state determination unit, and a response strategy unit. The similarity calculation unit is based on the cosine similarity calculation formula. : ; in, For standard translation vectors, Output the translated vector to the system; If the standard translation vector is: The system outputs the translated vector as follows: ,but: ; ; ; ; The translation strategy output module combines the revised comprehensive semantic vector with an authoritative dictionary database through multi-factor mapping rules to output translations that conform to the target language. This module not only optimizes the structure of the translation results but also adjusts the semantics of the translation based on context and cultural background, ensuring that the translated content conforms to the habits and understanding of the target language. It provides translations that are consistent with the target language's cultural background, making the translations more natural and fluent. The multi-factor mapping rules guarantee the semantic consistency and accuracy of the translation results, while the structural optimization of the translation results improves the readability and comprehensibility of the translated text.

[0051] The translation accuracy detection unit and the translation result automatic correction unit; The translation evaluation module consists of a similarity calculation unit, a status determination unit, and a response strategy unit. The similarity calculation unit uses the cosine similarity formula to calculate the similarity between the standard translation and the system-output translation. Then, combined with a credibility calculation function, it outputs a translation credibility level, categorized as high, medium, and low, to help the system assess translation quality. The status determination unit determines whether further correction is needed based on the credibility level of the translation result, while the response strategy unit provides correction suggestions. Based on cosine similarity and credibility calculations, it provides an accurate translation quality assessment. This multi-level translation evaluation helps quickly identify problem areas and make timely corrections. The automated evaluation function reduces the workload of manual review and improves system efficiency.

[0052] The translation accuracy detection unit evaluates the translation results output by the recognition module from multiple dimensions based on three indicators: semantic similarity, term matching degree, and contextual consistency, and sets a first accuracy threshold, a second accuracy threshold, and a third accuracy threshold. Combined with credibility calculation function : ; in, and To adjust the coefficients, the output translation credibility level is determined, categorized into high credibility, medium credibility, and low credibility. Semantic similarity, Terminology matching degree Context consistency.

[0053] Substitute the above calculated data into the credibility calculation function. Calculations can be performed to obtain... ; Therefore, the credibility of the translation is 0.951, which falls into the high credibility safe range.

[0054] When the semantic similarity is ≥0.90, the term matching degree is ≥0.85, and the contextual consistency is ≥0.88, the translation result is determined to be in the high accuracy safe range and no correction is required. When the semantic similarity is between 0.75 and 0.90, the term matching degree is between 0.65 and 0.85, and the contextual consistency is between 0.70 and 0.88, the translation result is determined to be in the warning range. The accuracy of the warning range is lower than that of the safe range, triggering the automatic correction unit to perform terminology standardization replacement and sentence optimization. When the semantic similarity is below 0.75, the term matching degree is below 0.65, or the contextual consistency is below 0.70, the translation result is determined to be in the danger zone. The accuracy of the danger zone is lower than that of the warning zone, and the system automatically starts multiple rounds of translation model iteration and manual verification prompts.

[0055] The automatic translation result correction unit includes a terminology standardization subunit, a grammatical structure adjustment subunit, and a human feedback fusion subunit. The terminology standardization subunit automatically replaces non-standard or easily confused professional terms in translation by calling a pre-set database of traditional Chinese medicine terminology, and ensures the rationality of the replacement through corpus verification. The grammar structure adjustment subunit uses natural language processing technology to optimize sentence structure and word order to ensure that the translation conforms to the expression habits of the target language. The human feedback fusion subunit is used to receive feedback from human experts on the correction results, and combines machine learning algorithms to dynamically update the translation model parameters and terminology database, forming a continuously iterative intelligent correction mechanism.

[0056] The state determination unit sets a safety threshold, a warning threshold, and a danger threshold based on the accuracy evaluation results output by the semantic correction module, and classifies the translation results into three categories: safe state, warning state, and critical state. The safety interval corresponds to an accuracy index ≥0.90, the translation results are stable, the system maintains the current translation model and terminology database, and no additional intervention is required; The warning range corresponds to an accuracy index between 0.75 and 0.90. The system will issue a warning, indicating that the translation result may have flaws and suggesting manual review or auxiliary correction. When a translation result has slight semantic deviations or non-standard terminology, the system categorizes it as a "warning zone." The system generates a "human review task list" for translations in the warning zone, automatically assigns them to the review pool, and sets up a 24-hour review feedback mechanism to prevent mistranslations or omissions. After human reviewers access the platform, they prioritize these results and provide correction suggestions based on the terminology database and corpus recommendation system. If a translation result is categorized as a warning zone due to contextual mismatch, the system automatically expands the context window, re-evaluates the semantic path, and further optimizes the sentence structure based on contextual consistency indicators to improve translation coherence and logical consistency. After each terminology replacement or sentence restructuring, the system re-executes three evaluation indicators to determine if it has entered the safe zone. If all indicators meet the standards after correction, the system marks it as "warning to safe" and records the correction path in the "correction evolution library" for subsequent system training.

[0057] When the accuracy index in the danger zone falls below 0.75, the system issues a danger alarm, initiates multi-model integrated translation and expert manual review to prevent the spread of erroneous translations.

[0058] When the system detects that a translation has serious semantic deviations, terminological confusion, and contextual breaks, it is interrupted as a non-standard expression or mistranslation and automatically determined to be in the "danger zone." For terminological errors or semantic incompleteness in the danger zone, the system activates a deep language understanding model to perform semantic layering, role labeling, and rewriting of the original sentence. All suspected erroneous terms or translation fragments are highlighted in red to indicate their danger level, and an error risk report is automatically generated. The report includes: original vocabulary, suspected mistranslation results, corresponding standard terminology recommendations, replacement suggestions, reference support, and semantic path analysis of the corpus. The system automatically triggers a high-priority task and pushes it to the Traditional Chinese Medicine translation expert platform, requiring review within 12 hours. After receiving the translation content, the expert review module can directly annotate and correct the mistranslated content. All translation entries corrected in the danger zone and their correction paths are finally archived in the "high-risk correction sample library." The system will periodically retrain the samples and optimize the future model's response strategy for high-risk semantic scenarios through reinforcement learning mechanisms.

[0059] The response strategy unit is used to output different strategies for each threshold range, including: When within the safe range, the system maintains the current translation algorithm and terminology database, and regularly updates and backtracks the quality of Chinese medicine professional corpus offline, without the need for immediate intervention; Within the safe zone, the system's primary strategy is to maintain the current version of the translation model and terminology database structure without making immediate adjustments. This approach is based on the following considerations: The high accuracy results indicate that current machine translation has strong generalization and expressive capabilities in processing traditional Chinese medicine terminology and sentence structures. The current Chinese-English correspondence and semantic vector matching degree in the terminology database have stabilized, indicating that the terminology standardization mechanism is working well and can cover the main processing scenarios of the system. It can effectively maintain the coherence between paragraphs and avoid context drift. Therefore, during system operation, once the translation result falls within the safe range, sub-processes such as terminology replacement, semantic correction, and structural optimization will not be triggered, ensuring that system resources are focused on other tasks to be corrected and improving overall response efficiency.

[0060] During the warning period, the system automatically triggers the semantic enhancement algorithm to perform secondary optimization translation based on the context, and automatically generates a manual review task to ensure that the review feedback cycle does not exceed 24 hours; In critical situations, the system immediately activates a multi-engine fusion translation strategy, combining rule matching and neural network deep learning technology to perform multiple rounds of correction, automatically label potential errors and high-risk terms, and generate a detailed correction suggestion report.

[0061] Specifically, the working process or principle of this AI-based cross-language assisted translation system for Traditional Chinese Medicine (TCM) is as follows: The core architecture of the system includes multiple modules. The data acquisition module is responsible for collecting relevant data on Chinese medicinal materials from multiple sources. This data includes lexical attributes, historical documents, regional differences, and pronunciation information. Through unified organization and standardization of this data, the system lays the foundation for subsequent semantic recognition and translation. The semantic recognition module extracts multi-dimensional semantic information of the names of Chinese medicinal materials using a deep learning model, including biological attribute factors and regional evolution factors. These factors help the system construct a multi-dimensional semantic network covering efficacy, uses, and cultural background, thereby achieving more accurate translation of Chinese medicinal material names. Based on this, the system's semantic correction module further corrects the identified semantics, using a historical evolution error function and a contextual ambiguity sensitivity model to eliminate potential translation errors. It also adjusts the translation results according to differences in different regions and historical backgrounds to ensure the accuracy and consistency of the translated content. After receiving the corrected semantic vector, the translation strategy output module combines the target language's grammatical rules and dictionary database to generate target language translations of the Chinese medicine names. In this process, the system not only focuses on grammatical accuracy but also optimizes the structure of the translation to ensure that the result conforms to the target language's expression habits while conveying the original meaning and cultural background. Furthermore, the system includes a translation evaluation module that scores the credibility of each translation and uses algorithms such as cosine similarity to assess translation quality, ensuring that the final output conforms to the professional standards and linguistic conventions of Chinese medicine names. After translation is completed, the system continuously optimizes the model through a feedback and learning module, using user feedback and expert proofreading results for incremental learning, gradually improving the accuracy and quality of the translation, thereby achieving self-learning and continuous improvement of the system. This system, by combining deep learning and semantic analysis technologies, not only provides strong support for the cross-language translation of Chinese medicine names, but also continuously improves translation quality through self-learning and feedback mechanisms, gradually adapting to the development needs of traditional Chinese medicine in the context of globalization. The system has broad application prospects and can provide accurate and efficient language services for the international development of traditional Chinese medicine, promoting the global dissemination and development of traditional Chinese medicine culture.

[0062] It should be noted that the power supply, specific composition and principle of the existing devices or equipment, or devices or equipment that can be implemented by existing technology, are clear to those skilled in the art, and therefore will not be described in detail.

Claims

1. An artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine, characterized in that, It includes a data acquisition module, a semantic recognition module, a semantic correction module, a translation strategy output module, and a translation evaluation module; The data acquisition module is used to collect multi-source data on the names of Chinese medicinal materials, including vocabulary attribute information, historical literature data, regional cultural background data and pronunciation features, and to perform multi-dimensional vector representation and normalization processing on the collected data. The semantic recognition module, based on natural language processing technology, constructs a semantic relationship network of Chinese medicinal herb names and extracts biological attribute factors from these names. Regional evolution factors and hierarchical semantic factors And establish a comprehensive semantic vector ; The semantic correction module is used to modify the comprehensive semantic vector. Combined with historical evolution error function Sensitivity coefficient for contextual ambiguity The correction factor is calculated, and the identification bias is corrected through the error accumulation function to obtain the corrected comprehensive semantic vector. ; The translation strategy output module is used to output the translation strategy based on the corrected comprehensive semantic vector. By combining dictionary databases and standard translation libraries, standard translation vectors are generated according to preset rules. Simultaneously, standard translation vectors are generated. Perform structural optimization and semantic matching adjustments; The translation evaluation module is used to evaluate the translation using a cosine similarity function. and credibility calculation function To generate target language translations The system performs a credibility score and outputs a classification of credibility levels, including high credibility, medium credibility, and low credibility, and provides corresponding correction suggestions.

2. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 1, characterized in that, The data acquisition module includes a vocabulary attribute extraction unit, a historical corpus integration unit, and a data preprocessing unit. The vocabulary attribute extraction unit obtains the pinyin, stroke count, part-of-speech tagging, and semantic labels of Chinese medicine names through natural language processing technology; The historical corpus integration unit integrates data from ancient books, modern documents, and folk terms for Chinese medicine names to form a time-series corpus database. The data preprocessing unit performs missing value filling, noise filtering, and multi-dimensional normalization on the collected multi-source data to ensure the accurate operation of the subsequent semantic recognition module.

3. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 2, characterized in that, The semantic recognition module includes a semantic inference unit and a semantic connection unit. The semantic inference unit adopts a semantic relation inference model based on graph neural networks to establish semantic connection weights between words. And calculate the nodes using the formula. The semantic vector representation; The semantic joint unit uses biological attribute factors Regional evolution factors and hierarchical semantic factors Joint encoding to form a comprehensive semantic vector .

4. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 3, characterized in that, The semantic correction module includes a semantic correction unit, a translation accuracy detection unit, and an automatic translation result correction unit. The semantic correction unit is used to utilize a historical evolution error function. ; And combined with the contextual ambiguity sensitivity coefficient We calculate the sensitivity of semantic vectors to contextual changes and adjust the translation strategy accordingly.

5. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 4, characterized in that, The translation strategy output module employs a function mapping model to generate standard translation vectors that conform to semantic and cultural context. .

6. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 5, characterized in that, The translation evaluation module includes a similarity calculation unit, a state determination unit, and a response strategy unit. The similarity calculation unit is based on the cosine similarity calculation formula. Calculate the standard translation vector With the system output translation vector And combined with the credibility calculation function Calculate the standard translation vector With the system output translation vector Credibility.

7. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 6, characterized in that, The translation accuracy detection unit and the translation result automatic correction unit; The translation accuracy detection unit is based on semantic similarity. Terminology matching degree Context consistency Three indicators are used to comprehensively evaluate the translation results output by the recognition module from multiple dimensions, and a first accuracy threshold, a second accuracy threshold, and a third accuracy threshold are set. When the semantic similarity is ≥0.90, the term matching degree is ≥0.85, and the contextual consistency is ≥0.88, the translation result is determined to be in the high accuracy safe range and no correction is required. When the semantic similarity is between 0.75 and 0.90, the term matching degree is between 0.65 and 0.85, and the contextual consistency is between 0.70 and 0.88, the translation result is determined to be in the warning range. The accuracy of the warning range is lower than that of the safe range, triggering the automatic correction unit to perform terminology standardization replacement and sentence optimization. When the semantic similarity is below 0.75, the term matching degree is below 0.65, or the contextual consistency is below 0.70, the translation result is determined to be in the danger zone. The accuracy of the danger zone is lower than that of the warning zone, and the system automatically starts multiple rounds of translation model iteration and manual verification prompts.

8. The artificial intelligence-based cross-language assisted translation system for Traditional Chinese Medicine according to claim 7, characterized in that, The automatic translation result correction unit includes a terminology standardization subunit, a grammatical structure adjustment subunit, and a human feedback fusion subunit. The terminology standardization subunit automatically replaces non-standard or easily confused professional terms in translation by calling a pre-set database of traditional Chinese medicine terminology, and ensures the rationality of the replacement through corpus verification. The grammar structure adjustment subunit uses natural language processing technology to optimize sentence structure and word order to ensure that the translation conforms to the expression habits of the target language. The human feedback fusion subunit is used to receive feedback from human experts on the correction results, and combines machine learning algorithms to dynamically update the translation model parameters and terminology database, forming a continuously iterative intelligent correction mechanism.

9. A cross-language assisted translation system for Traditional Chinese Medicine based on artificial intelligence according to claim 8, characterized in that, The state determination unit sets a safety threshold, a warning threshold, and a danger threshold based on the accuracy evaluation results output by the semantic correction module, and classifies the translation results into three categories: safe state, warning state, and critical state. The safety interval corresponds to an accuracy index ≥0.90, the translation results are stable, the system maintains the current translation model and terminology database, and no additional intervention is required; The warning range corresponds to an accuracy index between 0.75 and 0.

90. The system will issue a warning, indicating that the translation result may have flaws and suggesting manual review or auxiliary correction. When the accuracy index in the danger zone falls below 0.75, the system issues a danger alarm, initiates multi-model integrated translation and expert manual review to prevent the spread of erroneous translations.

10. A cross-language assisted translation system for Traditional Chinese Medicine based on artificial intelligence according to claim 9, characterized in that, The response strategy unit is used to output different strategies for each threshold range, including: When within the safe range, the system maintains the current translation algorithm and terminology database, and regularly updates and backtracks the quality of Chinese medicine professional corpus offline, without the need for immediate intervention; During the warning period, the system automatically triggers the semantic enhancement algorithm to perform secondary optimization translation based on the context, and automatically generates a manual review task to ensure that the review feedback cycle does not exceed 24 hours; In critical situations, the system immediately activates a multi-engine fusion translation strategy, combining rule matching and neural network deep learning technology to perform multiple rounds of correction, automatically label potential errors and high-risk terms, and generate a detailed correction suggestion report.

Images