Translation method and system for medical field and storage medium
By performing feature recognition and keyword extraction in the translation system, matching and comparing professional vocabulary in translated medical texts, the problem of difficulty in accurately translating medical terms in existing systems is solved, and high-quality medical translation is achieved.
Patent Information
- Application Number
- CN202411537847.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-31
- Publication Date
- 2025-03-11
AI Technical Summary
Existing translation systems are difficult to accurately translate terms in the medical field, and are prone to problems such as colloquialization of words and term replacement rather than inclusion in context, resulting in deviations in translation results.
By obtaining the medical text to be translated, feature recognition and keyword extraction are performed, the comparison translated text in the text database is matched based on the topic keywords, and professional vocabulary sentences are extracted and compared with the natural language processing algorithm.
It realizes professional translation of professional vocabulary and sentence patterns in medical texts, ensures the professionalism and accuracy of the translation, and reduces the deviation of translation results.
Smart Images
Figure CN119599033A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical translation technology, and in particular to a translation method, system and storage medium for the medical field. Background Art
[0002] The medical field involves a large number of professional terms, and the accurate translation of these terms is crucial to ensure the correct communication of medical information. If the translator does not have solid medical knowledge, he may feel at a loss when faced with difficult-to-understand medical articles. In addition, the breadth of the medical field requires translators to have a wide range of professional knowledge. If the knowledge is narrow, they will feel overwhelmed when translating. If traditional translation systems in existing technologies are used for translation, such as the Azure custom translation model, when using terms, it is easy to use colloquial terms, and it is only a term replacement rather than integrating the terms into the context. The final translated text is prone to deviations in translation results, and manual methods are needed to further adjust inappropriate wording.
[0003] Based on the above problems, there is an urgent need for a translation method that can translate medical texts in a way that meets the terminology requirements of the medical field. Summary of the invention In view of the fact that current translation systems are difficult to translate medical terms more appropriately, the present application provides a translation method, system and storage medium for the medical field to solve the above problem.
[0004] To achieve the above object, the present invention is implemented through the following technical solutions: The present application embodiment discloses a translation method for the medical field, including: Obtain medical texts to be translated; Perform feature recognition on the medical text to be translated and extract the medical text subject keywords based on the keyword extraction algorithm; Based on the subject keywords, the search method is used to match the translation texts in the text database that have been corrected and perfected in the specified language system; A pre-trained model based on natural language extraction technology extracts sentences with professional vocabulary from medical texts to be translated; Compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected translation text of the specified language system, extract the same words and sentences and translate them.
[0005] Adopting the above technical scheme: In the above scheme, it is possible to obtain texts of the same type that have been fully translated based on the keywords of the text to be translated, and to match the corresponding professional vocabulary based on the fully translated text, so as to realize professional translation of professional vocabulary and sentence patterns in medical texts. With a large amount of long-term training of the text database, it is possible to ensure the professionalism of medical text translation.
[0006] Preferably, for professional vocabulary sentences extracted by the natural language processing algorithm that do not correspond to the reference translation text of the corrected and perfected designated language system, the professional vocabulary sentences extracted by the natural language processing algorithm are translated through a medical field professional terminology database.
[0007] By adopting the above technical solution: for professional vocabulary sentences extracted by the natural language processing algorithm that do not correspond to the reference translation text of the corrected and perfected specified language system, this application can still achieve effective translation through the professional terminology database in the medical field to ensure the perfection of the translated text.
[0008] Further preferably, the method for extracting subject keywords of medical texts comprises: Extract medical content sentences from the medical text to be translated, perform word segmentation processing to obtain candidate word vectors and stop word vectors of the medical content sentences; Perform dependency syntactic analysis on the medical content sentence, and the matching correlation between all candidate word vectors in the medical content sentence and other candidate words in the medical content sentence; Analyze the occurrence frequency of the candidate word vectors in the extracted text and the degree of correlation between the candidate word vectors and other candidate words in the medical text to be translated, and obtain the degree of matching between the candidate word vectors and other candidate word vectors in the medical text to be translated; calculating the correlation weights of the candidate words in the medical content sentence and other candidate words in the medical content sentence; Establishing a weighted undirected graph for each medical content sentence, wherein the weighted undirected graph includes vertices for representing candidate words and weighted edges for representing relevant weights between candidate words; Through the text sorting algorithm, the score of each vertex of all weighted undirected graphs is calculated, the candidate words corresponding to the vertices are sorted according to the scores, and the keywords in the candidate words are extracted.
[0009] Adopting the above technical solution: The above solution can realize fast and effective extraction of special medical words and sentences in the medical text to be translated.
[0010] Further preferably, the retrieval method comprises: Obtain keywords, match keywords with synonyms, English abbreviations, and foreign language translations, and organize them into word sets; Sequentially retrieve a single keyword to all keywords in the word set and match them with the abstracts in the text database to obtain similar translated texts; Further match similar translation text content, count the probability of keyword repetition in the word set text, and use the text with the highest probability of repetition as the control translation text.
[0011] Adopting the above technical scheme: the above scheme can realize the retrieval and matching of the same type of text according to keywords, and can also match and search by synonyms, English abbreviations and foreign language translations of keywords, which can reduce the limitation that the matching model can only search according to the text itself. This retrieval model can match the keywords themselves and the synonyms, English abbreviations and foreign language translations of keywords, and can ensure the comprehensiveness of the matching.
[0012] Further preferably, the natural language processing algorithm includes: Perform word segmentation, word removal and sentence segmentation preprocessing on the text data to be processed, and remove stop words and punctuation marks; Use word frequency statistics to count the frequency of each word and phrase, and select words and phrases with high frequency as keywords; Build a word vector model and map each word or sentence into a high-dimensional vector space through a natural language processing model; Perform word vector addition and averaging operations on each word and sentence to obtain the vector representation of the word and sentence, use cosine similarity to calculate the similarity and compare the word and sentence similarities; Phrases with similarity higher than a threshold are regarded as identical phrases and extracted as translation phrases.
[0013] Adopting the above technical solution: the model in the above solution can match the medical words and sentences that need to be translated, and obtain the reference translation words and sentences that need to be translated into the specified language.
[0014] More preferably, it also includes a word and sentence proofreading method, and the word and sentence proofreading method includes: Manually review the translated text and extract inappropriately translated words and sentences; Correct the inappropriately translated words and sentences, and extract the inappropriately translated words and sentences and the corrected words and sentences; The model is trained by taking the inappropriately translated words and sentences as negative examples and the corrected words and sentences as positive examples.
[0015] Adopting the above technical scheme: the above scheme can further review the translated text, extract the inappropriately translated text, and re-translate it manually, use the inappropriately translated words and sentences as counter-examples, and use the corrected words and sentences as positive examples to train the model; it can avoid the inappropriate translation of the same type of words and sentences again, and can also achieve the perfection of the translation model.
[0016] Further preferably, a translation system for the medical field is applied to any of the translation methods described above, and is characterized by comprising: Feature extraction unit: extracts medical text subject keywords based on keyword extraction algorithm; Retrieval and matching unit: Based on the subject keywords, it matches the reference translation texts of the specified language system that have been corrected in the model database; Professional vocabulary extraction unit: extract professional vocabulary sentences based on the pre-processing training model of natural language extraction technology; Comparative translation unit: used to compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected comparative translation text of the specified language system, extract the same words and sentences and translate them; Further preferably, it also includes: Database translation unit: It uses a natural language processing algorithm to translate professional vocabulary sentences extracted from a database of professional terms in the medical field.
[0017] Further preferably, it also includes: Word and sentence proofreading unit: used to review the translated text, extract inappropriate text and correct it, use the inappropriately translated words and sentences as negative examples, and the corrected words and sentences as positive examples to train the model.
[0018] Further preferably, a computer-readable storage medium comprises a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned translation methods for the medical field. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0020] Figure 1 This is a flow chart of the translation method for the medical field in this application; Figure 2Flowchart for extracting medical text subject keywords for this application; Figure 3 A flowchart of the search method for this application; Figure 4 This is the natural language processing algorithm flow chart for this application; Figure 5 This is a flow chart of the word and sentence proofreading method for this application; Figure 6 This is a block diagram of the translation system for the medical field in this application.
[0021] In the figure: 1. Feature extraction unit; 2. Retrieval and matching unit; 3. Professional vocabulary extraction unit; 4. Comparative translation unit; 5. Database translation unit; 6. Word and sentence proofreading unit. DETAILED DESCRIPTION
[0022] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.
[0023] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.
[0024] See also Figure 1-6 For translation texts in the professional medical field, rigor and professionalism are often required. However, the current Azure custom translation model uses colloquial terms and only replaces terms instead of integrating them into the context. Based on the above problems, the present application discloses a translation method for the medical field, including: Obtain medical texts to be translated; Perform feature recognition on the medical text to be translated and extract the medical text subject keywords based on the keyword extraction algorithm; Based on the subject keywords, the search method is used to match the translation texts in the text database that have been corrected and perfected in the specified language system; A pre-trained model based on natural language extraction technology extracts sentences with professional vocabulary from medical texts to be translated; Compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected translation text of the specified language system, extract the same words and sentences and translate them. It is worth mentioning that the above scheme can achieve the acquisition of the same type of translated text based on the keywords of the text to be translated, and match the corresponding professional vocabulary based on the translated text, so as to achieve professional translation of professional vocabulary and sentence patterns in medical texts. With the large amount of long-term training of the text database, the professionalism of medical text translation can be guaranteed.
[0025] The solution provided in the above embodiment can match the medical vocabulary in the professional translation text with the text to be translated, so as to achieve a more professional translation. However, for the professional vocabulary sentences extracted by the natural language processing algorithm that do not correspond to the control translation text of the specified language system that has been corrected and improved, it may still be necessary to translate through the original translation model. Based on this, the embodiment of the present application translates the professional vocabulary sentences extracted by the natural language processing algorithm through the medical field professional term database. It is worth mentioning that for the professional vocabulary sentences extracted by the natural language processing algorithm that do not correspond to the control translation text of the specified language system that has been corrected and improved, the present application can still achieve effective translation through the medical field professional term database to ensure the perfection of the translated text.
[0026] The method for extracting medical text subject keywords comprises: Extract medical content sentences from the medical text to be translated, perform word segmentation processing to obtain candidate word vectors and stop word vectors of the medical content sentences; Perform dependency syntactic analysis on the medical content sentence, and the matching correlation between all candidate word vectors in the medical content sentence and other candidate words in the medical content sentence; Analyze the occurrence frequency of the candidate word vectors in the extracted text and the degree of correlation between the candidate word vectors and other candidate words in the medical text to be translated, and obtain the degree of matching between the candidate word vectors and other candidate word vectors in the medical text to be translated; calculating the correlation weights of the candidate words in the medical content sentence and other candidate words in the medical content sentence; Establishing a weighted undirected graph for each medical content sentence, wherein the weighted undirected graph includes vertices for representing candidate words and weighted edges for representing relevant weights between candidate words; Through the text sorting algorithm, the score of each vertex of all weighted undirected graphs is calculated, the candidate words corresponding to the vertex are sorted according to the score, and the keywords in the candidate words are extracted. The above scheme can realize the fast and effective extraction of special medical words and sentences in the medical text to be translated.
[0027] The retrieval method comprises: Obtain keywords, match keywords with synonyms, English abbreviations, and foreign language translations, and organize them into word sets; Sequentially retrieve a single keyword to all keywords in the word set and match them with the abstracts in the text database to obtain similar translated texts; Further match similar translated text content, count the probability of keyword repetition in the word set text, and use the text with the highest probability of repetition as the reference translation text. It is worth mentioning that the above scheme can not only search and match similar texts based on keywords, but also search by matching the synonyms, English abbreviations and foreign language translations of keywords, which can reduce the limitation of the matching model that can only search based on the text itself. This search model can match the keywords themselves and the synonyms, English abbreviations and foreign language translations of keywords, and can ensure the comprehensiveness of the match.
[0028] The natural language processing algorithm includes: Perform word segmentation, word removal and sentence segmentation preprocessing on the text data to be processed, and remove stop words and punctuation marks; Use word frequency statistics to count the frequency of each word and phrase, and select words and phrases with high frequency as keywords; Build a word vector model and map each word or sentence into a high-dimensional vector space through a natural language processing model; Perform word vector addition and averaging operations on each word and sentence to obtain the vector representation of the word and sentence, use cosine similarity to calculate the similarity and compare the word and sentence similarities; Phrases with similarity higher than a threshold are considered the same and extracted as translated phrases. The model in the above scheme can match medical phrases that need to be translated and obtain the reference translated phrases that need to be translated into a specified language.
[0029] Also included is a word and sentence proofreading method, the word and sentence proofreading method comprising: Manually review the translated text and extract inappropriately translated words and sentences; Correct the inappropriately translated words and sentences, and extract the inappropriately translated words and sentences and the corrected words and sentences; The model is trained by taking the inappropriately translated words and sentences as negative examples and the corrected words and sentences as positive examples.
[0030] It is worth mentioning that the above scheme can further review the translated text, extract the inappropriately translated text, and re-translate it manually, using the inappropriately translated words and sentences as counter-examples and the corrected words and sentences as positive examples to train the model; it can avoid the recurrence of inappropriate translation of the same type of words and sentences, and can also achieve the perfection of the translation model.
[0031] Further preferably, a translation system for the medical field is applied to any of the translation methods described above, and is characterized by comprising: Feature extraction unit 1: extracting medical text subject keywords based on keyword extraction algorithm; Retrieval and matching unit 2: Based on the subject keywords, matching the reference translation texts of the specified language system that have been corrected in the model database; Professional vocabulary extraction unit 3: extract professional vocabulary sentences based on the pre-processing training model of natural language extraction technology; Comparative translation unit 4: used to compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected and perfected comparative translation text of the specified language system, extract the same words and sentences and translate them; Further preferably, it also includes: Database translation unit 5: Translates the extracted professional vocabulary sentences through the medical field professional terminology database using a natural language processing algorithm.
[0032] Further preferably, it also includes: Word and sentence proofreading unit 6: used to review the translated text, extract inappropriate text and correct it, use the inappropriately translated words and sentences as negative examples, and use the corrected words and sentences as positive examples to train the model.
[0033] Further preferably, a computer-readable storage medium comprises a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned translation methods for the medical field.
[0034] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0035] Unless otherwise specified, the device components involved in the above embodiments are all conventional device components, and the connection methods and control methods involved are all conventional connection methods and control methods unless otherwise specified.
[0036] The present invention has been described in detail above in conjunction with the embodiments. However, those skilled in the art will appreciate that, without departing from the spirit of the present invention, the specific parameters in the above embodiments may be changed to form a plurality of specific embodiments, which are all within the common variation range of the present invention and will not be described in detail herein.
Claims
1. A translation method for the medical field, characterized in that: include: Obtain medical texts to be translated; Perform feature recognition on the medical text to be translated and extract the medical text subject keywords based on the keyword extraction algorithm; Based on the subject keywords, the search method is used to match the translation texts in the text database that have been corrected and perfected in the specified language system; A pre-trained model based on natural language extraction technology extracts sentences with professional vocabulary from medical texts to be translated; Compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected translation text of the specified language system, extract the same words and sentences and translate them.
2. A translation method for the medical field according to claim 1, characterized in that: For professional vocabulary sentences extracted by the natural language processing algorithm that do not correspond to the reference translation text of the corrected and perfected designated language system, the professional vocabulary sentences extracted by the natural language processing algorithm are translated through the medical field professional terminology database.
3. A translation method for the medical field according to claim 1, characterized in that: The method for extracting medical text subject keywords comprises: Extract medical content sentences from the medical text to be translated, perform word segmentation processing to obtain candidate word vectors and stop word vectors of the medical content sentences; Perform dependency syntactic analysis on the medical content sentence, and the matching correlation between all candidate word vectors in the medical content sentence and other candidate words in the medical content sentence; Analyze the occurrence frequency of the candidate word vectors in the extracted text and the degree of correlation between the candidate word vectors and other candidate words in the medical text to be translated, and obtain the degree of matching between the candidate word vectors and other candidate word vectors in the medical text to be translated; calculating the correlation weights of the candidate words in the medical content sentence and other candidate words in the medical content sentence; Establishing a weighted undirected graph for each medical content sentence, wherein the weighted undirected graph includes vertices for representing candidate words and weighted edges for representing relevant weights between candidate words; Through the text sorting algorithm, the score of each vertex of all weighted undirected graphs is calculated, the candidate words corresponding to the vertices are sorted according to the scores, and the keywords in the candidate words are extracted.
4. A translation method for the medical field according to claim 1, characterized in that: The retrieval method comprises: Obtain keywords, match keywords with synonyms, English abbreviations, and foreign language translations, and organize them into word sets; Sequentially retrieve a single keyword to all keywords in the word set and match them with the abstracts in the text database to obtain similar translated texts; Further match similar translation text content, count the probability of keyword repetition in the word set text, and use the text with the highest probability of repetition as the control translation text.
5. A translation method for the medical field according to claim 1, characterized in that: The natural language processing algorithm includes: Perform word segmentation, word removal and sentence segmentation preprocessing on the text data to be processed, and remove stop words and punctuation marks; Use word frequency statistics to count the frequency of each word and phrase, and select words and phrases with high frequency as keywords; Build a word vector model and map each word or sentence into a high-dimensional vector space through a natural language processing model; Perform word vector addition and averaging operations on each word and sentence to obtain the vector representation of the word and sentence, use cosine similarity to calculate the similarity and compare the word and sentence similarities; Phrases with similarity higher than a threshold are regarded as identical phrases and extracted as translation phrases.
6. A translation method for the medical field according to claim 1, characterized in that: Also included is a word and sentence proofreading method, the word and sentence proofreading method comprising: Manually review the translated text and extract inappropriately translated words and sentences; Correct the inappropriately translated words and sentences, and extract the inappropriately translated words and sentences and the corrected words and sentences; The model is trained by taking the inappropriately translated words and sentences as negative examples and the corrected words and sentences as positive examples.
7. A translation system for the medical field, applied to the translation method according to any one of claims 1 to 6, characterized in that: include: Feature extraction unit: extracts medical text subject keywords based on keyword extraction algorithm; Retrieval and matching unit: Based on the subject keywords, it matches the reference translation texts of the specified language system that have been corrected in the model database; Professional vocabulary extraction unit: extract professional vocabulary sentences based on the pre-processing training model of natural language extraction technology; Comparative translation unit: used to compare the professional vocabulary and sentences extracted by the natural language processing algorithm with the corrected comparative translation text of the specified language system, extract the same words and sentences and translate them.
8. A translation system for the medical field according to claim 7, characterized in that: Also includes: Database translation unit: It uses a natural language processing algorithm to translate professional vocabulary sentences extracted from a database of professional terms in the medical field.
9. A translation system for the medical field according to claim 7, characterized in that: Also includes: Word and sentence proofreading unit: used to review the translated text, extract inappropriate text and correct it, use the inappropriately translated words and sentences as negative examples, and the corrected words and sentences as positive examples to train the model.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the translation method for the medical field according to any one of claims 1 to 6.