Text translation optimization analysis system and method based on artificial intelligence

Abstract

The invention discloses a text translation optimization analysis system and method based on artificial intelligence, and relates to the technical field of text translation optimizing.The method comprises the steps that the translation key degree of candidate professional terms for professional texts is evaluated, and target professional terms are obtained; performing context consistency and text adaptation degree analysis on the candidate terminology translations, and evaluating the cross-language similarity degree of the candidate terminology translations and the target terminology to obtain target terminology translations; performing text translation on the professional text to obtain a translated text of the professional text, and optimizing the translated text to obtain a target translated text; according to the method, the target translation text of the professional text is acquired, the target professional term translations of different target professional terms in the professional text are acquired, the translation output data of the professional text are generated, and the translation output data are output to a user through a translation platform, so that the accuracy of professional text translation is improved, and the problem of mistranslation of the professional terms in the professional text is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of text translation optimization, and specifically to a text translation optimization analysis system and method based on artificial intelligence. Background Art

[0002] With the continuous advancement of the globalization process, the frequency of cross-language text communication in both work and study is increasing day by day. Especially for texts in professional fields, different professional fields have their own unique terms. An incorrect translation of a professional term may very likely cause the entire content of the text to change, resulting in errors in the text content. Moreover, the concepts in texts related to professional fields are often defined very strictly and have complex logical relationships. Excellent translation can ensure that the concepts in the text are accurately conveyed. At the same time, a professional text translation with precise wording and compliance with norms can also avoid unnecessary losses caused by communication ambiguities.

[0003] For the traditional translation of texts in professional fields, different databases are very likely to be distributed in different projects or documents, lacking unified management of data. As a result, accurate translation content for professional terms in the text cannot be found in the existing databases, and manual translation will also lead to multiple translation methods for the same professional term, reducing the consistency of translation. In addition, the existing translation tools cannot solve the problem of incorrect translation of professional terms and have poor adaptability to complex text structures. Summary of the Invention

[0004] The purpose of the present invention is to provide a text translation optimization analysis system and method based on artificial intelligence to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A text translation optimization analysis method based on artificial intelligence, the method comprising: Step S100: Obtain the professional text uploaded by the user on the translation platform, perform word segmentation on the professional text to obtain candidate professional terms of the professional text, and evaluate the key degree of the candidate professional terms for the translation of the professional text to obtain target professional terms; Step S200: Obtain candidate professional term translations of the target professional terms from the translation platform, obtain the paragraphs where the target professional terms are located in the professional text, analyze the context consistency and text adaptability of the candidate professional term translations, and evaluate the cross-language similarity degree between the candidate professional term translations and the target professional terms to obtain target professional term translations; Step S300: Obtain the target professional term translations of the target professional terms, perform text translation on the professional text to obtain a translation text of the professional text, and optimize the translation text to obtain a target translation text; Step S400: Obtain the target translation text of the professional text, obtain the target professional term translations of different target professional terms in the professional text, generate the translation output data of the professional text, and output the translation output data to the user through the translation platform.

[0006] Further, step S100 includes: Step S101: Obtain the professional text uploaded by the user from the translation platform, perform word segmentation on the professional text, and preprocess the word-segmented professional text to obtain each keyword of the professional text; Step S102: Through the preset regular expressions in the translation platform, screen out candidate professional terms from each keyword, obtain each candidate professional term of the professional text, and evaluate the key degree of the candidate professional terms for the translation of the professional text. The specific evaluation process is as follows: Obtain the frequency A of the candidate professional term appearing in the professional text sum , obtain the total number N of each keyword in the professional text, obtain the corpus of the field to which the professional text belongs from the translation platform, and calculate the document key value F of the candidate professional term: , B sum is the total number of documents in the corpus; B´ sum is the total number of documents in the corpus that contain the candidate professional term; Step S103: Calculate the term ambiguity value α of the candidate professional term: , where d is the total number of translation candidates corresponding to the candidate professional term obtained from the translation platform; Step S104: Obtain the position of the candidate professional term in the professional text, obtain the position importance factor corresponding to different positions from the translation platform, and obtain the position importance factor β of the candidate professional term according to the position of the candidate professional term in the professional text; Step S105: Normalize the document key value F, term ambiguity value α, and position importance factor β of the candidate professional term, and calculate the key score H of the candidate professional term: , where γ1, γ2, and γ3 are the preset first weight coefficient, second weight coefficient, and third weight coefficient respectively, γ1, γ2, and γ3 are all greater than 0, and γ1 + γ2 + γ3 = 1; When the key score H is greater than the preset key score threshold, it is determined that the candidate professional term plays a key role in the translation of the professional text, and the candidate professional term is recorded as the target professional term of the professional text, and each target professional term of the professional text is obtained; In the above steps, since professional texts in a specific field contain many professional terms, some of these terms are simple for the translation model and have no impact on the translation of professional texts. If these professional terms are also optimized in translation, it will undoubtedly greatly increase the burden on the translation platform and affect the performance of the translation platform. Therefore, obtaining the key professional terms in the professional text can not only ensure the accuracy of translation but also greatly reduce the load of the translation platform during the translation process, further improving the performance of the translation platform.

[0007] Further, step S200 includes: Step S201: Obtain the target professional terms of the professional text, obtain the translation database in the translation platform, and obtain several candidate professional term translations of the target professional terms from the translation database; Step S202: Obtain the paragraph where the target professional term is located from the professional text and record it as the target paragraph of the target professional term; Perform a context consistency analysis on the candidate professional term translations of the target professional term. The specific analysis process is as follows: Obtain the sentence containing the target professional term from the target paragraph and record it as the target sentence of the target professional term. Use the candidate professional term translation as the translation of the target professional term, translate the target sentence, and use a preset language model to obtain the vector v of the candidate professional term translation in the target sentence τ ; Obtain each sentence containing the candidate professional term translation from the translation database, and calculate the context consistency score R of the candidate professional term translation: , where v △ is the average value of the vectors of the candidate professional term translation in each sentence obtained using the language model; Step S203: Perform a text fitness analysis on the candidate professional term translations of the target professional term. The specific analysis process is as follows: Obtain the corpus of each field from the translation platform, obtain the mean value G of the occurrence frequencies of the candidate professional term translation in each document in the corpus of the field to which the professional text belongs, and obtain the mean value G of the total number of words contained in each document sum , calculate the domain frequency P of the candidate professional term translation G =G / G sum ; Obtain the mean value M of the occurrence frequencies of the candidate professional term translation in several documents in the corpus of each field, and obtain the mean value M of the total number of words contained in several documents in the corpus of each field sum , calculate the general frequency P of the candidate professional term translation M =M / Msum ; Calculate the text fitness P of the candidate professional term translation: P = log(M / M sum ); Step S204: Evaluate the cross - language similarity between the candidate professional term translation and the target professional term. The specific evaluation process is as follows: Use a pre - set cross - language embedding model to map the candidate professional term translation and the target professional term into the same vector space, and obtain the vector E' of the candidate professional term translation and the vector E of the target professional term respectively △ , calculate the cosine similarity between the vector E' and the vector E △ , and obtain the cross - language similarity value E between the candidate professional term translation and the target professional term; Step S205: Obtain the text fitness, context consistency score, and cross - language similarity value of several candidate professional term translations of the target professional term and perform normalization processing; Calculate the translation score Y of the candidate professional term translation: Y = η E ×E + η P ×P + η R ×R, where η E is the weight coefficient corresponding to the cross - language similarity value E, η P is the weight coefficient corresponding to the text fitness P, and η R is the weight coefficient corresponding to the context consistency score R; Obtain the translation scores of several candidate professional term translations. When the translation score of a candidate professional term translation is the maximum among the translation scores of several candidate professional term translations, record the candidate professional term translation as the target professional term translation of the target professional term.

[0008] Furthermore, step S300 includes: Step S301: Obtain the target professional term translations corresponding to each target professional term in the professional text, and perform text translation on the professional text in the translation platform to obtain the translated text of the professional text; Step S302: Obtain the paragraph translations corresponding to each paragraph in the professional text, collect the paragraph translations corresponding to each paragraph to obtain the paragraph translation set of the professional text, and use a pre - set language model to perform vector transformation on the paragraph translations corresponding to each paragraph in the paragraph translation set to obtain the semantic vectors corresponding to the paragraph translations of each paragraph; Calculate the paragraph translation consistency L of the translated text: , where j is the total number of each paragraph; u i is the semantic vector corresponding to the paragraph translation of the i - th paragraph in the paragraph translation set; u i+1It is the semantic vector corresponding to the paragraph translation of the (i + 1)-th paragraph in the paragraph translation set; Step S303: When the paragraph translation consistency L is greater than the preset paragraph translation consistency threshold L´, it is determined that the translation text does not need to be optimized, and the translation text is recorded as the target translation text of the professional text; When L ≤ L´, it is determined that the translation text needs to be optimized for translation. The specific process of optimizing the translation text is as follows: Calculate the average value of the cosine similarity between the semantic vectors corresponding to the paragraph translations of each paragraph in the paragraph translation set and the semantic vectors corresponding to the paragraph translations of adjacent paragraphs, and record it as the paragraph consistency value of each paragraph; When the paragraph consistency value of a certain paragraph is less than the preset paragraph consistency threshold, optimize the translation of a certain paragraph. Then, obtain several candidate professional terms for a certain target professional term in a certain paragraph, and sequentially replace the target professional term translation of a certain target professional term according to the translation scores of the several candidate professional terms until the paragraph consistency value of a certain paragraph is greater than or equal to the preset paragraph consistency threshold. When the paragraph consistency value of a certain paragraph is greater than or equal to the preset paragraph consistency threshold, obtain a certain candidate professional term corresponding to a certain target professional term, and use a certain candidate professional term as the target professional term translation of a certain target professional term; Step S304: Optimize the translations of several paragraphs in the paragraph translation set whose paragraph consistency values are less than the preset paragraph consistency threshold until L > L´, and record the translation text as the target translation text of the professional text.

[0009] Furthermore, step S400 includes: Step S401: Obtain the target translation text of the professional text in the translation platform, obtain the target professional term translations of each target professional term in the professional text, and collect them according to the corresponding relationship to obtain the term replacement report of the professional text; Step S402: Collect the professional text, the target translation text, and the term replacement report to obtain the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform.

[0010] In order to better implement the above method, a text translation optimization analysis system based on artificial intelligence is also proposed. The system includes a term key degree evaluation module, a translation analysis module, a translation optimization module, and a translation output module; The term key degree evaluation module is used to obtain the professional text uploaded by the user in the translation platform, obtain the candidate professional terms of the professional text, and evaluate the key degree of the candidate professional terms for the translation of the professional text to obtain the target professional terms; A translation analysis module for obtaining candidate professional term translations of target professional terms, analyzing the context consistency and text adaptability of the candidate professional term translations, and evaluating the cross - language similarity between the candidate professional term translations and the target professional terms to obtain the target professional term translations; A translation optimization module for text - translating the professional text according to the target professional term translations of the target professional terms to obtain a translation text, and optimizing the translation text to obtain the target translation text; A translation output module for obtaining the target professional term translations of different target professional terms in the professional text, combining with the target translation text, generating translation output data of the professional text, and outputting the translation output data to the user through a translation platform.

[0011] Furthermore, the term criticality evaluation module includes a candidate term acquisition unit and a term criticality evaluation unit; The candidate term acquisition unit is used to obtain the professional text uploaded by the user, perform word - segmentation processing on the professional text to obtain each keyword of the professional text, and obtain the candidate professional terms of the professional text from each keyword; The term criticality evaluation unit is used to evaluate the translation criticality of the candidate professional terms to the professional text to obtain the target professional terms.

[0012] Furthermore, the translation analysis module includes a candidate translation acquisition unit and a translation analysis unit; The candidate translation acquisition unit is used to obtain the target professional terms of the professional text and obtain the candidate professional term translations of the target professional terms from the translation platform; The translation analysis unit is used to analyze the context consistency and text adaptability of the candidate professional term translations, and evaluate the cross - language similarity between the candidate professional term translations and the target professional terms to obtain the target professional term translations.

[0013] Furthermore, the translation optimization module includes a text translation unit and a translation optimization unit; The text translation unit is used to obtain the target professional term translations of the target professional terms and perform text translation on the professional text to obtain the translation text of the professional text; The translation optimization unit is used to optimize the translation text to obtain the target translation text.

[0014] Furthermore, the translation output module includes a translation output unit; The translation output unit is used to obtain the target translation text of the professional text in the translation platform, obtain the term replacement report of the professional text, generate the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention realizes accurate and efficient translation of texts in professional fields. Considering various keywords contained in professional texts and different translation difficulties of different keywords, not only are professional terms obtained from professional texts, but also the key degree of different professional terms for professional text translation is analyzed, so as to obtain target professional terms that play a key role in professional text translation. By using means of context consistency and text adaptation degree, the translation effects of different candidate professional term translations of the target professional terms are analyzed, and the best candidate professional term translation is selected as the target professional term translation of the target professional terms, and the professional text is translated. According to the translation situation, the translated text of the professional text is optimized, so as to improve the accuracy of professional text translation and solve the problem of mistranslation of professional terms in professional texts. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a method flow chart of a method for optimizing and analyzing text translation based on artificial intelligence according to the present invention; Figure 2 is a module schematic diagram of a system for optimizing and analyzing text translation based on artificial intelligence according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution, a method for optimizing and analyzing text translation based on artificial intelligence, and the method includes: Step S100: Obtain the professional text uploaded by the user on the translation platform, perform word segmentation processing on the professional text, obtain candidate professional terms of the professional text, and evaluate the key degree of the candidate professional terms for the translation of the professional text to obtain target professional terms; Among them, step S100 includes: Step S101: Obtain the professional text uploaded by the user from the translation platform, perform word segmentation processing on the professional text, and perform preprocessing on the word-segmented professional text to obtain each keyword of the professional text; The preprocessing includes removing stop words and restoring the word form; Step S102: Through the regular expressions preset in the translation platform, screen out candidate professional terms from each keyword, obtain each candidate professional term of the professional text, and evaluate the key degree of the candidate professional term for the translation of the professional text. The specific evaluation process is as follows: Obtain the frequency A of the candidate professional term appearing in the professional text sum , obtain the total number N of each keyword in the professional text, obtain the corpus of the field to which the professional text belongs from the translation platform, and calculate the document key value F of the candidate professional term: , B sum is the total number of documents in the corpus; B´ sum is the total number of documents containing the candidate professional term in the corpus; For example, the regular expression for matching the field of pharmaceutical chemistry in the medical field is: , Its function is to identify chemical formulas. For example, CO2 will be matched when using this regular expression; Step S103: Calculate the term ambiguity value α of the candidate professional term: , where d is the total number of translation candidates corresponding to the candidate professional term obtained from the translation platform; Step S104: Obtain the position of the candidate professional term in the professional text, obtain the position importance factor corresponding to different positions from the translation platform, and obtain the position importance factor β of the candidate professional term according to the position of the candidate professional term in the professional text; For example, different position importance factors corresponding to different positions in the text. Generally, the importance factor is set to 0.8 in the title of the text, 0.5 in the abstract, and 0.3 in the main body; Step S105: Normalize the document key value F, term ambiguity value α, and position importance factor β of the candidate professional term, and calculate the key score H of the candidate professional term: , where γ1, γ2, and γ3 are the preset first weight coefficient, second weight coefficient, and third weight coefficient respectively. γ1, γ2, and γ3 are all greater than 0, and γ1 + γ2 + γ3 = 1; For example, γ1, γ2, and γ3 are 0.4, 0.4, and 0.2 respectively; the document key value F, term ambiguity value α, and position importance factor β of the candidate professional term are 0.7, 0.8, and 0.6 respectively; Calculate the key score H of the candidate professional term: , When the key score H is greater than the preset key score threshold, it is determined that the candidate professional term plays a key role in the translation of the professional text. Then, the candidate professional term is recorded as the target professional term of the professional text, and each target professional term of the professional text is obtained. Step S200: Obtain the candidate professional term translations of the target professional term from the translation platform, obtain the paragraph where the target professional term is located in the professional text, analyze the context consistency and text adaptability of the candidate professional term translations, and evaluate the cross - language similarity between the candidate professional term translations and the target professional term to obtain the target professional term translations. Among them, step S200 includes: Step S201: Obtain the target professional terms of the professional text, obtain the translation database in the translation platform, and obtain several candidate professional term translations of the target professional term from the translation database. For example, the translation database is a database used to store, manage, and retrieve language resources, containing translations and documents corresponding to different professional terms. Step S202: Obtain the paragraph where the target professional term is located from the professional text and record it as the target paragraph of the target professional term. Perform context consistency analysis on the candidate professional term translations of the target professional term. The specific analysis process is as follows: Obtain the sentence containing the target professional term from the target paragraph and record it as the target sentence of the target professional term. Use the candidate professional term translation as the translation of the target professional term to translate the target sentence, and use the preset language model to obtain the vector v of the candidate professional term translation in the target sentence. τ ; For example, the language model is the BERT model. BERT is an innovative pre - trained language processing model launched by Google in 2018. It is built based on the powerful Transformer architecture. By introducing advanced technologies such as bidirectional context understanding and masked language model (MLM), it can greatly improve the performance in various natural language processing (NLP) tasks, making the machine more accurate and efficient in understanding and generating human language. For example, for the acquisition process of the vector v τ is that when the candidate professional term translation is τ and the target sentence is S, then v τ =BERT(s)[index(τ)], where BERT(s) is the vector matrix output by the BERT language model for the target sentence, and index(τ) represents the position index of the candidate professional term translation τ in the target sentence S. When the professional term translation τ is a multi - word term, then v τ is the position of the first word in the position index in the target sentence S. Obtain each sentence containing candidate translations of professional terms from the translation database, and calculate the context consistency score R of the candidate translations of professional terms: , where v △ is the average value of the vectors of the candidate translations of professional terms obtained using the language model in each sentence; Step S203: Conduct a text fitness analysis on the candidate translations of the target professional terms. The specific analysis process is as follows: Obtain the corpora of each field from the translation platform, obtain the average value G of the occurrence frequencies of the candidate translations of professional terms in each document in the corpus of the field to which the professional text belongs, and obtain the average value G of the total number of words contained in each document sum , and calculate the field frequency P of the candidate translations of professional terms G =G / G sum ; For example, the corpus is a database collected and organized by the translation platform, including various documents in the field to which it belongs; For example, G sum is 2000, G is 20, and calculate the field frequency P of the candidate translations of professional terms G =20 / 2000 = 1 / 100 = 0.01; Obtain the average value M of the occurrence frequencies of the candidate translations of professional terms in several documents in the corpora of each field, and obtain the average value M of the total number of words contained in several documents in the corpora of each field sum , and calculate the general frequency P of the candidate translations of professional terms M =M / M sum ; Calculate the text fitness P of the candidate translations of professional terms = log(M / M sum ); Step S204: Evaluate the cross - language similarity between the candidate translations of professional terms and the target professional terms. The specific evaluation process is as follows: Use a preset cross - language embedding model to map the candidate translations of professional terms and the target professional terms into the same vector space, and respectively obtain the vector E´ of the candidate translations of professional terms and the vector E of the target professional terms △ , and calculate the cosine similarity between the vector E´ and the vector E △ , to obtain the cross - language similarity value E between the candidate translations of professional terms and the target professional terms; For example, the cross - language embedding model is the LaBSE model. The LaBSE model is a multilingual sentence embedding model developed by Google, supporting up to 109 languages. By combining the pre - training techniques of masked language modeling (MLM) and translation language modeling (TLM), this model can efficiently generate cross - language sentence embeddings, thus enabling semantic understanding and comparison between different languages; Step S205: Obtain the text fitness, context consistency score, and cross - language similarity value of several candidate translations of the target technical term, and perform normalization processing; Calculate the translation score Y of the candidate translation of the technical term = η E ×E + η P ×P + η R ×R, where η E is the weight coefficient corresponding to the cross - language similarity value E, η P is the weight coefficient corresponding to the text fitness P, and η R is the weight coefficient corresponding to the context consistency score R; For example, η E + η P + η R = 1, and η E , η P and η R are often adjusted according to task requirements. In professional translation, context consistency and domain fitness may be more important. It can be set that η R = 0.4, η P = 0.4, η E = 0.2; Obtain the translation scores of several candidate translations of the technical term. When the translation score of a candidate translation of the technical term is the maximum among the translation scores of several candidate translations of the technical term, record the candidate translation as the target translation of the target technical term; Step S300: Obtain the target translation of the target technical term, perform text translation on the professional text to obtain the translated text of the professional text, and optimize the translated text to obtain the target translated text; Among them, step S300 includes: Step S301: Obtain the target translations corresponding to each target technical term of the professional text, and perform text translation on the professional text in the translation platform to obtain the translated text of the professional text; Step S302: Obtain the paragraph translations corresponding to each paragraph in the professional text, collect the paragraph translations corresponding to each paragraph to obtain the paragraph translation set of the professional text, and use a preset language model to perform vector transformation on the paragraph translations corresponding to each paragraph in the paragraph translation set to obtain the semantic vectors corresponding to the paragraph translations of each paragraph; Calculate the paragraph translation consistency L of the translated text: , where j is the total number of each paragraph; u i is the semantic vector corresponding to the paragraph translation of the i-th paragraph in the paragraph translation set; u i+1 is the semantic vector corresponding to the paragraph translation of the (i + 1)-th paragraph in the paragraph translation set; Step S303: When the paragraph translation consistency L is greater than the preset paragraph translation consistency threshold L´, it is determined that the translated text does not need to be optimized, and the translated text is recorded as the target translated text of the professional text; When L ≤ L´, it is determined that the translated text needs to be optimized for translation. The specific process of optimizing the translated text is as follows: Calculate the average value of the cosine similarity between the semantic vectors corresponding to the paragraph translations of each paragraph in the paragraph translation set and the semantic vectors corresponding to the paragraph translations of adjacent paragraphs, and record it as the paragraph consistency value of each paragraph; When the paragraph consistency value of a certain paragraph is less than the preset paragraph consistency threshold, optimize the translation of a certain paragraph, then obtain several candidate professional terms for a certain target professional term in a certain paragraph, and replace the target professional term translation of a certain target professional term in turn according to the translation scores of several candidate professional terms until the paragraph consistency value of a certain paragraph is greater than or equal to the preset paragraph consistency threshold, and obtain a certain candidate professional term corresponding to a certain target professional term when the paragraph consistency value of a certain paragraph is greater than or equal to the preset paragraph consistency threshold, and use a certain candidate professional term as the target professional term translation of a certain target professional term; Step S304: Optimize the translations of several paragraphs in the paragraph translation set whose paragraph consistency values are less than the preset paragraph consistency threshold until L > L´, and record the translated text as the target translated text of the professional text; Step S400: Obtain the target translated text of the professional text, obtain the target professional term translations of different target professional terms in the professional text, generate the translation output data of the professional text, and output the translation output data to the user through the translation platform; Among them, Step S400 includes: Step S401: Obtain the target translated text of the professional text in the translation platform, obtain the target professional term translations of each target professional term in the professional text, and collect them according to the corresponding relationship to obtain the term replacement report of the professional text; Step S402: Collect the professional text, the target translated text and the term replacement report to obtain the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform; To better implement the above method, a text translation optimization analysis system based on artificial intelligence is also proposed. The system includes a term key degree evaluation module, a translation analysis module, a translation optimization module, and a translation output module; The term key degree evaluation module is used to obtain the professional text uploaded by the user on the translation platform, obtain the candidate professional terms of the professional text, evaluate the key degree of the candidate professional terms for the translation of the professional text, and obtain the target professional terms; The translation analysis module is used to obtain the candidate professional term translations of the target professional terms, analyze the context consistency and text fitness of the candidate professional term translations, and evaluate the cross - language similarity degree between the candidate professional term translations and the target professional terms, and obtain the target professional term translations; The translation optimization module is used to perform text translation on the professional text according to the target professional term translations of the target professional terms, obtain the translation text, and optimize the translation text to obtain the target translation text; The translation output module is used to obtain the target professional term translations of different target professional terms in the professional text, combine them with the target translation text, generate the translation output data of the professional text, and output the translation output data to the user through the translation platform; Among them, the term key degree evaluation module includes a candidate term acquisition unit and a term key degree evaluation unit; The candidate term acquisition unit is used to obtain the professional text uploaded by the user, perform word segmentation on the professional text to obtain each keyword of the professional text, and obtain the candidate professional terms of the professional text from each keyword; The term key degree evaluation unit is used to evaluate the key degree of the candidate professional terms for the translation of the professional text and obtain the target professional terms; Among them, the translation analysis module includes a candidate translation acquisition unit and a translation analysis unit; The candidate translation acquisition unit is used to obtain the target professional terms of the professional text and obtain the candidate professional term translations of the target professional terms from the translation platform; The translation analysis unit is used to analyze the context consistency and text fitness of the candidate professional term translations, and evaluate the cross - language similarity degree between the candidate professional term translations and the target professional terms, and obtain the target professional term translations; Among them, the translation optimization module includes a text translation unit and a translation optimization unit; The text translation unit is used to obtain the target professional term translations of the target professional terms and perform text translation on the professional text to obtain the translation text of the professional text; The translation optimization unit is used to optimize the translation text to obtain the target translation text; Among them, the translation output module includes a translation output unit; The translation output unit is configured to obtain the target translation text of the professional text in the translation platform, obtain the term replacement report of the professional text, generate the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform.

[0019] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced by the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. An artificial intelligence-based method for optimizing and analyzing text translation, characterized in that, The method includes: Step S100: Obtain the professional text uploaded by the user on the translation platform, perform word segmentation on the professional text to obtain candidate professional terms of the professional text, evaluate the key degree of the candidate professional terms for translating the professional text, and obtain target professional terms; Step S200: Obtain candidate professional term translations of the target professional terms from the translation platform, obtain the paragraphs where the target professional terms are located in the professional text, analyze the context consistency and text fitness of the candidate professional term translations, and evaluate the cross - language similarity between the candidate professional term translations and the target professional terms to obtain target professional term translations; Step S300: Obtain the target professional term translations of the target professional terms, perform text translation on the professional text to obtain a translation text of the professional text, and optimize the translation text to obtain a target translation text; Step S400: Obtain the target translation text of the professional text, obtain the target professional term translations of different target professional terms in the professional text, generate translation output data of the professional text, and output the translation output data to the user through the translation platform.

2. The method for optimizing and analyzing text translation based on artificial intelligence according to claim 1, characterized in that The step S100 includes: Step S101: Obtain the professional text uploaded by the user from the translation platform, perform word segmentation on the professional text, and pre - process the word - segmented professional text to obtain each keyword of the professional text; Step S102: Through a preset regular expression in the translation platform, screen out candidate professional terms from each of the keywords, obtain each candidate professional term of the professional text, and evaluate the key degree of the candidate professional terms for translating the professional text. The specific evaluation process is as follows: Obtain the frequency A of the candidate professional term appearing in the professional text sum , obtain the total number N of each keyword in the professional text, obtain the corpus of the field to which the professional text belongs from the translation platform, and calculate the document key value F of the candidate professional term: ， B sum To obtain the total number of documents in the corpus; B' sum To obtain the total number of documents in the corpus that contain the candidate technical term; Step S103: Calculate the term ambiguity value α of the candidate professional term: ， where d is the total number of translation candidates corresponding to the candidate professional term obtained from the translation platform; Step S104: Obtain the position where the candidate professional term is located in the professional text, obtain the position importance factors corresponding to different positions from the translation platform, and obtain the position importance factor β of the candidate professional term according to the position where the candidate professional term is located in the professional text; Step S105: Perform normalization processing on the document key value F, term ambiguity value α, and position importance factor β of the candidate professional term, and calculate the key score H of the candidate professional term: ， where γ1, γ2, and γ3 are respectively preset first weight coefficient, second weight coefficient, and third weight coefficient, γ1, γ2, and γ3 are all greater than 0, and γ1 + γ2 + γ3 = 1; When the key score H is greater than a preset key score threshold, it is determined that the candidate professional term plays a key role in translating the professional text, then the candidate professional term is recorded as the target professional term of the professional text, and each target professional term of the professional text is obtained.

3. The method for optimizing and analyzing text translation based on artificial intelligence according to claim 2, wherein, The step S200 includes: Step S201: Obtain the target professional terms of the professional text, obtain the translation database in the translation platform, and obtain several candidate professional term translations of the target professional terms from the translation database; Step S202: Obtain the paragraph in which the target professional term is located from the professional text, and record it as the target paragraph of the target professional term; Perform a context consistency analysis on the candidate professional term translations of the target professional term. The specific analysis process is as follows: Obtain the sentence containing the target technical term from the target paragraph, and denote it as the target sentence of the target technical term. Use the candidate technical term translation as the translation of the target technical term, translate the target sentence, and use a preset language model to obtain the vector v of the candidate technical term translation in the target sentence τ ; Obtain each sentence containing the candidate professional term translation from the translation database, and calculate the context consistency score R of the candidate professional term translation; ， where v △ is the average value of the vectors of the candidate professional term translations obtained using the language model in each of the sentences; Step S203: Perform a text fitness analysis on the candidate professional term translations of the target professional term. The specific analysis process is as follows: Obtain the corpus in each field from the translation platform, obtain the average value G of the frequency of occurrence of the candidate professional term translation in each document in the corpus of the field to which the professional text belongs, and obtain the average value G of the total number of words contained in each document sum , calculate the field frequency P of the candidate professional term translation G =G / G sum ; Obtain the mean value M of the frequencies of occurrences of the candidate professional term translation in several documents in the corpora of each field, and obtain the mean value M of the total number of words contained in several documents in the corpora of each field sum , calculate the general frequency P of the candidate professional term translation M = M / M sum ; Calculate the text fitness P of the candidate professional term translation as P = log(M / M sum ); Step S204: Evaluate the cross-language similarity between the candidate professional term translations and the target professional term. The specific evaluation process is as follows: Using a preset cross - language embedding model, map the candidate professional term translation and the target professional term to the same vector space, and respectively obtain the vector E´ of the candidate professional term translation and the vector E of the target professional term △ , calculate the cosine similarity between the vector E´ and the vector E △ , and obtain the cross - language similarity value E between the candidate professional term translation and the target professional term; Step S205: Obtain the text fitness, context consistency score, and cross-language similarity values of several candidate professional term translations of the target professional term and perform normalization processing; Calculate the translation score Y of the candidate professional term translation as Y = η E × E + η P × P + η R × R, where η E is the weight coefficient corresponding to the cross - language similarity value E, η P is the weight coefficient corresponding to the text fitness P, η R is the weight coefficient corresponding to the context consistency score R; Obtain the translation scores of the several candidate professional term translations. When the translation score of a candidate professional term translation is the maximum among the several candidate professional term translations, record the candidate professional term translation as the target professional term translation of the target professional term.

4. An optimization analysis method for text translation based on artificial intelligence according to claim 3, characterized in that The said step S300 includes: Step S301: Obtain the target professional term translations corresponding to the respective target professional terms of the professional text, perform text translation on the professional text in the translation platform to obtain the translation text of the professional text; Step S302: Obtain the paragraph translations corresponding to the respective paragraphs in the professional text, pool the paragraph translations corresponding to the respective paragraphs to obtain the paragraph translation set of the professional text, and use a preset language model to perform vector transformation on the paragraph translations corresponding to the respective paragraphs in the paragraph translation set to obtain the semantic vectors corresponding to the paragraph translations of the respective paragraphs; Calculate the paragraph translation consistency L of the translation text; ， where j is the total number of the respective paragraphs; u i is the semantic vector corresponding to the paragraph translation of the i-th paragraph in the paragraph translation set; u i+1 is the semantic vector corresponding to the paragraph translation of the (i + 1)-th paragraph in the paragraph translation set; Step S303: When the paragraph translation consistency L is greater than the preset paragraph translation consistency threshold L´, it is determined that the translation text does not need to be optimized, and then record the translation text as the target translation text of the professional text; When L ≤ L´, it is determined that the translation text needs to be translation-optimized. The specific process of optimizing the translation text is as follows: Calculate the average value of the cosine similarities between the semantic vectors corresponding to the paragraph translations of the respective paragraphs in the paragraph translation set and the semantic vectors corresponding to the paragraph translations of the adjacent paragraphs, and record it as the paragraph consistency value of the respective paragraphs; When the paragraph consistency value of a certain paragraph is less than the preset paragraph consistency threshold, translation optimization is performed on the certain paragraph. Then, several candidate professional terms of a certain target professional term in the certain paragraph are obtained, and the target professional term translation of the certain target professional term is sequentially replaced according to the translation scores of the several candidate professional terms until the paragraph consistency value of the certain paragraph is greater than or equal to the preset paragraph consistency threshold. When the paragraph consistency value of a certain paragraph is greater than or equal to the preset paragraph consistency threshold, a certain candidate professional term corresponding to a certain target professional term is obtained, and the certain candidate professional term is used as the target professional term translation of the certain target professional term; Step S304: Perform translation optimization on several paragraphs in the paragraph translation set whose paragraph consistency values are less than the preset paragraph consistency threshold until L > L´, and then record the translation text as the target translation text of the professional text.

5. The method for optimizing and analyzing text translation based on artificial intelligence according to claim 4, wherein The step S400 includes: Step S401: Obtain the target translation text of the professional text in the translation platform, obtain the target professional term translations of each target professional term in the professional text, and collect them according to the corresponding relationship to obtain the term replacement report of the professional text; Step S402: Collect the professional text, the target translation text, and the term replacement report to obtain the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform.

6. An artificial intelligence-based text translation optimization analysis system for performing an artificial intelligence-based text translation optimization analysis method according to any one of claims 1-5, characterized in that, The system includes a term criticality assessment module, a translation analysis module, a translation optimization module, and a translation output module; The term criticality assessment module is used to obtain the professional text uploaded by the user in the translation platform, obtain the candidate professional terms of the professional text, and evaluate the translation criticality of the candidate professional terms for the professional text to obtain the target professional terms; The translation analysis module is used to obtain the candidate professional term translations of the target professional terms, analyze the context consistency and text fitness of the candidate professional term translations, and evaluate the cross - language similarity degree between the candidate professional term translations and the target professional terms to obtain the target professional term translations; The translation optimization module is used to perform text translation on the professional text according to the target professional term translations of the target professional terms to obtain a translation text, and optimize the translation text to obtain the target translation text; The translation output module is used to obtain the target professional term translations of different target professional terms in the professional text, and combine them with the target translation text to generate the translation output data of the professional text, and output the translation output data to the user through the translation platform.

7. An artificial intelligence-based text translation optimization analysis system according to claim 6, characterized in that, The term criticality assessment module includes a candidate term acquisition unit and a term criticality assessment unit; The candidate term acquisition unit is used to acquire the professional text uploaded by the user, perform word segmentation on the professional text to obtain each keyword of the professional text, and acquire candidate professional terms of the professional text from the various keywords; The term key degree evaluation unit is used to evaluate the key degree of the candidate professional terms for the translation of the professional text to obtain target professional terms.

8. An artificial intelligence-based text translation optimization analysis system according to claim 6, characterized in that, The translation analysis module includes a candidate translation acquisition unit and a translation analysis unit; The candidate translation acquisition unit is used to acquire the target professional terms of the professional text and acquire candidate professional term translations of the target professional terms from the translation platform; The translation analysis unit is used to analyze the context consistency and text fitness of the candidate professional term translations, and evaluate the cross-language similarity degree between the candidate professional term translations and the target professional terms to obtain target professional term translations.

9. An artificial intelligence-based text translation optimization analysis system according to claim 6, characterized in that The translation optimization module includes a text translation unit and a translation optimization unit; The text translation unit is used to acquire the target professional term translations of the target professional terms and perform text translation on the professional text to obtain the translation text of the professional text; The translation optimization unit is used to optimize the translation text to obtain the target translation text.

10. An artificial intelligence-based text translation optimization analysis system according to claim 6, characterized in that, The translation output module includes a translation output unit; The translation output unit is used to acquire the target translation text of the professional text in the translation platform, acquire the term replacement report of the professional text, generate the translation output data of the professional text, and output the translation output data to the user of the professional text through the translation platform.

Images