Method and system for translating terminology vocabularies in ship field by LLM based on qLoRA fine tuning

Abstract

The invention relates to the technical field of ship translation, in particular to a method and system for translating terminology vocabularies in the ship field through LLM based on qLoRA fine tuning. Comprising the following steps: intelligently collecting and establishing a ship field professional data lexicon through a trainer, performing fine tuning on a large language model by virtue of a qLoRA fine tuning technology, and converting related data of the ship field professional data lexicon into SQL statements by utilizing the fine-tuned large language model; obtaining a to-be-translated text, converting the to-be-translated text into an SQL statement, carrying out matching analysis on the SQL statement and the ship domain database, and screening and calibrating the terminology text of the ship domain; based on a ship field professional data word library, matching corresponding professional vocabulary groups for professional terms calibrated by the to-be-translated text; through a set translation matching algorithm, the scheme can effectively translate professional vocabularies in the ship field.

Description

Technical Field

[0001] The present invention relates to the technical field of ship translation, and specifically relates to a method and system for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM. Background Art

[0002] Intelligent translation refers to the process of using artificial intelligence technology for language translation. It uses specific computer programs to translate natural languages in one written form, image, or sound form into another form of natural language, involving comprehensive technologies of text, speech, and images, also known as multimodal machine translation. The working principle of an intelligent translation system mainly includes two steps: language processing and machine translation. In the language processing stage, the system analyzes and processes the input original text, including word segmentation, part-of-speech tagging, syntactic analysis, etc., with the aim of converting the original text into a form that can be understood and processed by a computer. In the machine translation stage, the system matches and compares the text processed by language processing with a pre-trained language model, and then generates a translation result in the target language according to the prediction result of the model. However, in the field of ship technology, there are a large number of professional vocabulary, which are extremely important for the information transmission in the ship technology field. For some professional vocabulary, the system will not perform corresponding and compliant matching according to the ship technology field scenario, resulting in inaccurate translation.

[0003] Based on the above problems, there is an urgent need for a method and system for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM, which can solve the problem of ineffective translation of professional terms. Summary of the Invention Aiming at the inaccurate translation of the corresponding professional vocabulary in the ship field by the current translation system, this application is provided to solve the above problems.

[0004] To achieve the above object, the present invention is realized through the following technical solutions: An embodiment of this application discloses a method for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM, which is characterized in that a professional data word library in the ship field is intelligently collected and established by a trainer, the large language model is fine-tuned relying on qLoRA fine-tuning technology, and the relevant data in the professional data word library in the ship field is converted into SQL statements by using the fine-tuned large prediction model; S1: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the ship field database, and screen and calibrate the professional term text in the ship field; S2: Based on the professional data word library in the ship field, match the corresponding professional vocabulary groups for the professional terms calibrated in the text to be translated; S3: Based on the sentence patterns where the professional vocabulary in the text to be translated is located, calculate the matching rates of the professional vocabulary groups through the set translation matching algorithm, and screen out the vocabulary with the highest matching rate. S4: Perform a full-text translation on the text to be translated, load the vocabulary with the highest matching rate into the text with partial translation completed, and complete the translation of the entire text.

[0005] Adopting the above technical solution: The above solution provides a method that can screen and calibrate the professional vocabulary in the text specialized in the ship field, calculate the matching rates, calculate the vocabulary with the highest matching rate, and use the vocabulary with the highest matching rate as the vocabulary for professional translation, which can effectively translate the patent text in the ship field.

[0006] Preferably, the said S3 includes: S31: Based on the screened professional technical text, analyze the sub-field of the text in the ship industry, match it with the sub-database of the ship industry, match the corresponding sub-field database, match the screened vocabulary in the professional vocabulary group with the sub-database, obtain the field matching factor, and use the field matching factor as the calculation parameter of the translation matching algorithm. S32: Load the vocabulary of the professional vocabulary group into the original text and translated text comparison text respectively. Based on the translated text, analyze the coherence of the text sentence meaning through the sentence meaning translation algorithm, and obtain the coherence factor of the text. S33: According to the sub-field of the text in the ship industry, screen out the original text and translated text comparison text with high matching rate in the original text and translated text database, match similar sentence patterns based on the translated text, and find the corresponding professional vocabulary in the translated text to obtain the historical matching factor.

[0007] Adopting the above technical solution: The above solution further discloses step S3, which can analyze and obtain the matching degree factors to be considered in the translation matching algorithm, including the field matching factor, the coherence factor, and the historical matching factor, and analyze the matching degree of the professional vocabulary based on the above factors.

[0008] Further preferably, the establishment method of the sub-database of the sub-field of the ship industry includes: S311: Confirm the sub-field types of the ship industry, and establish different sub-professional vocabulary sub-libraries according to the sub-field types. S312: Based on different sub-professional vocabulary sub-libraries, calculate the matching rates of the screened vocabulary in the professional vocabulary group differently, and sum up the matching rates of different sub-professional vocabulary sub-libraries to obtain the final matching factor.

[0009] Adopt the above technical solution: The above solution can identify the sub - field types in the shipbuilding industry, and calculate the matching rate based on the sub - professional vocabulary sub - library, and calculate the final matching factor.

[0010] Further preferably, the analysis method of the coherence factor includes: S321: Identify the basic components of the translated text, analyze the disorder rate of the sentence components after translation, and use the disorder rate parameter of the sentence components as the first consideration factor; S322: Identify and judge the word order of the translated text, judge the disorder rate of the word order of the sentence after translation, and use the disorder rate of the word order as the second consideration factor; S323: Reverse - translate the translated text to obtain the reverse - translated text, calculate the dissimilarity rate between the reverse - translated text and the text to be translated, and use the dissimilarity rate of the text to be translated as the third consideration factor; S324: Calculate the final text coherence factor through the coherence algorithm based on the third consideration factor, the second consideration factor and the first consideration factor.

[0011] Adopt the above technical solution: The above solution further details the calculation of the text coherence factor, and can calculate the final coherence factor based on the disorder rate of sentence components, the word order of the text, and the dissimilarity rate of reverse translation.

[0012] Further preferably, after the translation of all the text is completed, professional technicians process and proofread the translated text; Extract the wrong words and wrong word orders in the translated text, and conduct corresponding proofreading; Modify and polish the relatively mechanical sentences in the translation, and replace the original text with the modified and polished sentences; Load the translated text processed by professional technicians and the original text into the original - translation text database.

[0013] Adopt the above technical solution: After the system translates the text, the above solution then extracts the wrong word orders manually, modifies and polishes the mechanical sentences, and loads the finally translated text into the original - translation text database.

[0014] Further preferably, the formula of the translation matching algorithm is: , Where, is the matching rate, is the domain matching influence coefficient, Sub - professional vocabulary sub - library matching rate, is the number of sub - fields, is the third consideration factor, is the first consideration factor, is the second consideration factor, is the text coherence influence coefficient; is the historical matching factor, is the historical matching influence coefficient.

[0015] Adopt the above technical solution: The above solution further discloses the formula of the translation matching algorithm, and calculates the matching rate of professional vocabulary based on the matching rate of the professional vocabulary sub-library, the text coherence influence factor, and the historical matching factor.

[0016] Further preferably, the S33 includes: S331: Extract the professional vocabulary in the text that can reflect the text theme; S332: Analyze the professional vocabulary that reflects the text theme, determine the text theme, and screen out all the original-translation comparison copies that match the text theme in the original-translation text database; S333: Perform sentence pattern matching on all the original-translation comparison copies that match the text theme, and match the original-translation comparison copy with the highest sentence pattern overlap; S334: Based on the translated text, match similar sentence patterns, find the corresponding professional vocabulary in the translated text, and obtain the historical matching factor.

[0017] Adopt the above technical solution: The above solution further refines the solution of step S33, and can realize matching similar sentence patterns based on the translated text, finding the corresponding professional vocabulary in the translated text, and calculating the historical matching factor.

[0018] A system is applied to a method for translating professional term vocabulary in the ship field based on a qLoRA fine-tuned LLM as described in any one of the above, including: Screening and calibration unit: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the ship field database, and screen and calibrate the professional term text in the ship field; Professional vocabulary matching unit: Based on the professional data word library in the ship field, match the corresponding professional vocabulary groups for the professional terms calibrated in the text to be translated; Matching rate calculation unit: Through the set translation matching algorithm, calculate the matching rate of the professional vocabulary group based on the sentence pattern where the professional vocabulary in the text to be translated is located, and screen out the vocabulary with the highest matching rate; Full text translation unit: Perform full text translation on the text to be translated, and load the vocabulary with the highest matching rate into the text with partial translation completed to complete the translation of all texts.

[0019] Further preferably, the matching rate calculation unit includes: Field matching factor analysis module: Based on the selected professional technical texts, analyze the sub - fields of the texts in the shipbuilding industry, match them with the sub - database of the shipbuilding industry, match the corresponding sub - database, match the selected vocabulary in the professional vocabulary group with the sub - database, obtain the field matching factor, and use the field matching factor as a calculation parameter for the translation matching algorithm; Text coherence factor analysis module: Load the vocabulary of the professional vocabulary group into the original - translation comparison text respectively. Based on the translated text, analyze the coherence of the text meaning through the sentence - meaning translation algorithm to obtain the text coherence factor; Historical matching factor analysis module: According to the sub - fields of the text in the shipbuilding industry, screen the original - translation comparison texts with high matching rates in the original - translation text database. Based on the translated text, match similar sentence patterns and find the corresponding professional vocabulary in the translated text to obtain the historical matching factor.

[0020] Further preferably, the text coherence factor analysis module includes: First consideration factor analysis module: Identify the basic components of the translated text, analyze the disorder rate of the sentence components of the translated sentence, and use the disorder rate parameter of the sentence components as the first consideration factor; Second consideration factor analysis module: Identify and judge the word order of the translated text, judge the disorder rate of the word order of the translated sentence, and use the disorder rate of the word order as the second consideration factor; Third consideration factor analysis module: Reverse - translate the translated text to obtain the reverse - translated text, calculate the dissimilarity rate between the reverse - translated text and the text to be translated, and use the dissimilarity rate of the text to be translated as the third consideration factor; Coherence factor calculation module: Based on the third consideration factor, the second consideration factor and the first consideration factor, calculate the final text coherence factor through the coherence algorithm. Description of the Drawings

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 Is the flow chart of the method for translating professional - term vocabulary of the present application; Figure 2 For the present application Figure 1 Is the flow chart of step S3 in the present application; Figure 3 For this application Figure 2 Detailed flowchart of step S31 in this application; Figure 4 For this application Figure 2 Detailed flowchart of step S32 in this application. Specific implementation manners

[0023] In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of this application. However, those skilled in the art should clearly understand that this application can 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 avoid unnecessary details from interfering with the description of this application.

[0024] It should be understood that when used in the specification of this application and the appended claims, the term "comprising" indicates the presence of the 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 their combinations.

[0025] Please refer to Figures 1-4 , like the original translation system device, when translating vocabulary in the field of shipbuilding, it may not be able to match professional vocabulary that conforms to the field of the sentence, resulting in inaccurate expression of the final intended meaning. Based on the above problems, the embodiments of this application disclose a method for translating professional terms in the field of shipbuilding based on qLoRA fine-tuning of LLM, which is characterized in that a professional data vocabulary library in the field of shipbuilding is intelligently collected and established by a trainer, the large language model is fine-tuned relying on the qLoRA fine-tuning technology, and the relevant data in the professional data vocabulary library in the field of shipbuilding is converted into SQL statements by the fine-tuned large prediction model; S1: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the shipbuilding field database, and screen and calibrate the professional term text in the shipbuilding field; S2: Based on the professional data vocabulary library in the field of shipbuilding, match the corresponding professional vocabulary groups for the professional terms calibrated in the text to be translated; S3: Through the set translation matching algorithm, calculate the matching rate of the professional vocabulary groups based on the sentence pattern where the professional vocabulary in the text to be translated is located, and screen out the vocabulary with the highest matching rate; S4: Perform full-text translation on the text to be translated, load the vocabulary with the highest matching rate into the text with partial translation completed, and complete the translation of the entire text. It is worth mentioning that the above solution provides a method to screen and calibrate the professional vocabulary in the text of the ship field, calculate the matching rate, calculate the vocabulary with the highest matching rate, and use the vocabulary with the highest matching rate as the vocabulary for professional translation, which can effectively translate the patent text in the ship field.

[0026] Preferably, the above S3 includes: S31: Based on the screened professional technical text, analyze the sub-field of the text in the ship industry, match it with the sub-database of the ship industry, match the corresponding sub-field database, match the selected vocabulary in the professional vocabulary group with the sub-database, obtain the field matching factor, and use the field matching factor as the calculation parameter of the translation matching algorithm; this step can realize the matching judgment of the sub-field of the text to be translated, and finally determine the sub-field of the text in the ship industry. For example, for transportation ships, ocean development ships, fishing boats, engineering ships, cruise ships, and warships, there are corresponding vocabulary with higher matching rates in different fields. Screening the sub-field in advance can also ensure the matching speed of the translated text.

[0027] S32: Load the vocabulary of the professional vocabulary group into the original text-translation comparison text respectively. Based on the translated text, analyze the coherence of the text sentence meaning through the sentence meaning translation algorithm to obtain the coherence factor of the text; after the system translates the text, although the translation of the entire sentence pattern is completed, for the language expression methods of different countries, after the text is translated and compared, the position of the translated text is not necessarily accurate. Therefore, it is necessary to consider the coherence of the text sentence meaning.

[0028] S33: According to the sub-field of the text in the ship industry, screen the original text-translation comparison text with a high matching rate in the original text-translation text database, match similar sentence patterns based on the translated text, and find the corresponding professional vocabulary in the translated text to obtain the historical matching factor; in the originally translated text, the corresponding vocabulary has been screened and confirmed. Screening the original text-translation comparison text with a high matching rate in the original text-translation text database can screen out the vocabulary that meets the requirements to a great extent.

[0029] It is worth mentioning that the above solution further discloses step S3, which can analyze and obtain the matching degree factors to be considered in the translation matching algorithm, including the field matching factor, the coherence factor, and the historical matching factor, and analyze the matching degree of professional vocabulary based on the above factors.

[0030] The establishment method of the sub-database of the sub-field of the ship industry includes: S311: Confirm the types of sub - fields in the shipbuilding industry and establish different sub - libraries of specialized vocabulary according to the types of sub - fields; S312: Based on different sub - libraries of specialized vocabulary, calculate the matching rates of the selected words in the group of specialized vocabulary differently, and sum up the matching rates of different sub - libraries of specialized vocabulary to obtain the final matching factor.

[0031] It is worth mentioning that the above - mentioned solution can confirm the types of sub - fields in the shipbuilding industry and calculate the matching rate based on the sub - library of specialized vocabulary to calculate the final matching factor.

[0032] The analysis method of the coherence factor includes: S321: Identify the basic components of the translated text, analyze the disorder rate of the sentence components in the translated sentence, and use the disorder rate parameter of the sentence components as the first consideration factor; S322: Identify and judge the word order of the translated text, judge the disorder rate of the word order of the translated sentence, and use the disorder rate of the word order as the second consideration factor; S323: Reverse - translate the translated text to obtain the reverse - translated text, calculate the dissimilarity rate between the reverse - translated text and the text to be translated, and use the dissimilarity rate of the text to be translated as the third consideration factor; S324: Calculate the final text coherence factor through a coherence algorithm based on the third consideration factor, the second consideration factor and the first consideration factor.

[0033] It is worth mentioning that the above - mentioned solution further details the calculation of the text coherence factor and can calculate the final coherence factor based on the disorder rate of sentence components, the word order of the text, and the dissimilarity rate of reverse translation.

[0034] Further preferably, after the translation of all the text is completed, professional technical personnel process and proofread the translated text; Extract the wrong words and wrong word orders in the translated text and conduct corresponding proofreading; Modify and polish the sentences with relatively mechanical translations, and replace the original text with the modified and polished sentences; Load the translated text processed by professional technical personnel and the original text into the original - translation text database.

[0035] It is worth mentioning that the above - mentioned solution, after the text is translated by a systematic solution, then extracts the wrong word orders manually, modifies and polishes the mechanical sentences, and loads the finally translated text into the original - translation text database.

[0036] Further preferably, the formula of the translation matching algorithm is: , Among them, is the matching rate, is the domain matching influence coefficient, the matching rate of the sub-library of specialized vocabulary, is the number of sub-domains, is the third consideration factor, is the first consideration factor, is the second consideration factor, is the text coherence influence coefficient; is the historical matching factor, is the historical matching influence coefficient.

[0037] It is worth mentioning that: The above scheme further discloses the formula of the translation matching algorithm, and calculates the matching rate of professional vocabulary based on the matching rate of the sub-library of professional vocabulary, the text coherence influence factor, and the historical matching factor. The above formula is based on the professional vocabulary selected above, and according to the matching rate of the professional vocabulary in the sub-library of professional vocabulary in the sentence, text coherence, and historical matching factors, etc., different influence coefficients are assigned to the three factors, and finally the matching rates of different vocabulary are obtained, and the vocabulary with high matching rate is selected as the finally determined translation vocabulary.

[0038] The said S33 includes: S331: Extract the professional vocabulary in the text that can reflect the text theme; S332: Analyze the professional vocabulary that reflects the text theme, determine the text theme, and screen out all the original text-translation counterparts that match the text theme in the original text-translation database; S333: Among all the original text-translation counterparts that match the text theme screened out, perform sentence pattern matching to match the original text-translation counterpart with the highest sentence pattern overlap; S334: Based on the translated text, match similar sentence patterns, find the corresponding professional vocabulary in the translated text, and obtain the historical matching factor.

[0039] It is worth mentioning that the above scheme further refines the scheme of step S33, and can realize matching similar sentence patterns based on the translated text, finding the corresponding professional vocabulary in the translated text, and calculating the historical matching factor.

[0040] A system, applied to a method for translating professional term vocabulary in the ship domain based on qLoRA fine-tuning of LLM as described in any one of the above, includes: Screening and calibration unit: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the ship domain database, and screen and calibrate the professional term text in the ship domain; Professional vocabulary matching unit: Based on the professional data word library in the ship field, match the professional terms marked in the text to be translated with the corresponding professional vocabulary groups; Matching rate calculation unit: Through the set translation matching algorithm, calculate the matching rate of the professional vocabulary group based on the sentence pattern where the professional vocabulary in the text to be translated is located, and screen out the vocabulary with the highest matching rate; Full-text translation unit: Perform full-text translation on the text to be translated, load the vocabulary with the highest matching rate into the text with partial translation completed, and complete the translation of the entire text.

[0041] The said matching rate calculation unit includes: Field matching factor analysis module: Based on the screened professional term text, analyze the subdivision field of the text in the ship industry, match it with the ship industry subdivision database, match the corresponding subdivision database, match the screened vocabulary in the professional vocabulary group with the subdivision database, obtain the field matching factor, and use the field matching factor as the calculation parameter of the translation matching algorithm; Text coherence factor analysis module: Load the vocabulary of the professional vocabulary group into the original text and translated text comparison text respectively. Based on the translated text, analyze the coherence of the text meaning through the sentence meaning translation algorithm, and obtain the text coherence factor; Historical matching factor analysis module: According to the subdivision field of the text in the ship industry, screen out the original text and translated text comparison text with high matching rate in the original text and translated text database. Based on the translated text, match similar sentence patterns, and find the corresponding professional vocabulary in the translated text to obtain the historical matching factor.

[0042] The said text coherence factor analysis module includes: First consideration factor analysis module: Identify the basic components of the translated text, analyze the disorder rate of the sentence components of the translated sentence, and use the disorder rate parameter of the sentence components as the first consideration factor; Second consideration factor analysis module: Identify and judge the word order of the translated text, judge the disorder rate of the word order of the translated sentence, and use the disorder rate of the word order as the second consideration factor; Third consideration factor analysis module: Reverse translate the translated text to obtain the reverse translated text, calculate the dissimilarity rate between the reverse translated text and the text to be translated, and use the dissimilarity rate of the text to be translated as the third consideration factor; Coherence factor calculation module: Based on the third consideration factor, the second consideration factor and the first consideration factor, calculate the final text coherence factor through the coherence algorithm.

[0043] In the above embodiments, the device components involved are all conventional device components unless otherwise specified. The connection methods and control methods involved are all conventional connection methods and control methods unless otherwise specified.

[0044] The present invention has been described in detail above in conjunction with the embodiments. However, those skilled in the art can understand that, without departing from the gist of the present invention, various specific parameters in the above embodiments can be changed to form multiple specific embodiments, which are all within the common variation range of the present invention and will not be elaborated herein one by one.

Claims

1. A method for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM, characterized in that, Including: Intelligently collect and establish a professional data thesaurus for the ship field through a trainer, fine-tune a large language model relying on the qLoRA fine-tuning technology, and use the fine-tuned large prediction model to convert the relevant data in the professional data thesaurus for the ship field into SQL statements; S1: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the ship field database, and screen and calibrate the professional term texts in the ship field; S2: Based on the professional data thesaurus for the ship field, match the calibrated professional terms in the text to be translated with the corresponding professional vocabulary groups; S3: Through the set translation matching algorithm, calculate the matching rate of the professional vocabulary groups based on the sentence pattern where the professional vocabulary in the text to be translated is located, and screen out the vocabulary with the highest matching rate; S4: Perform full-text translation on the text to be translated, load the vocabulary with the highest matching rate into the text with partial translation completed, and complete the translation of all texts.

2. A method for translating professional terminology in the ship field using an LLM fine-tuned based on qLoRA, characterized in that The above S3 includes: S31: Based on the screened professional term texts, analyze the sub-fields of the text in the ship industry, match with the sub-databases of the ship industry, match the corresponding sub-field databases, match the screened vocabulary in the professional vocabulary groups with the sub-databases, obtain the field matching factor, and use the field matching factor as the calculation parameter of the translation matching algorithm; S32: Load the vocabulary of the professional vocabulary groups into the original text-translation comparison text respectively. Based on the translated text, analyze the coherence of the text meaning through the sentence meaning translation algorithm, and obtain the text coherence factor; S33: According to the sub-fields of the text in the ship industry, screen the original text-translation comparison texts with high matching rates in the original text-translation text database, match similar sentence patterns based on the translated text, and find the corresponding professional vocabulary in the translated text to obtain the historical matching factor.

3. The method for translating professional terminology in the ship field by an LLM fine-tuned based on qLoRA according to claim 2, wherein The establishment method of the sub-database of the sub-fields of the ship industry includes: S311: Confirm the types of sub-fields of the ship industry, and establish different sub-professional vocabulary sub-libraries according to the types of sub-fields; S312: Based on different sub-professional vocabulary sub-libraries, calculate the matching rates of the screened vocabulary in the professional vocabulary groups differently, and sum up the matching rates of different sub-professional vocabulary sub-libraries to obtain the final matching factor.

4. A method for translating professional term vocabulary in the ship field by an LLM fine-tuned based on qLoRA, characterized in that, The analysis method of the above coherence factor includes: S321: Identify the basic components of the translated text, analyze the disorder rate of the sentence components of the translated sentence, and use the disorder rate parameter of the sentence components as the first consideration factor; S322: Identify and judge the word order of the translated text, judge the disorder rate of the word order of the translated sentence, and use the disorder rate of the word order as the second consideration factor; S323: Reverse translate the translated text to obtain the reverse translation text, calculate the difference rate between the reverse translation text and the text to be translated, and use the difference rate of the text to be translated as the third consideration factor; S324: Based on the third consideration factor, the second consideration factor, and the first consideration factor, calculate the final text coherence factor through the coherence algorithm.

5. A method for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM according to claim 2, characterized in that After completing the translation of all texts, professional technicians process and proofread the translated texts. Extract the wrong words and incorrect word orders in the translated texts and conduct corresponding proofreading. Modify and polish the relatively mechanical sentences in the translation, and replace the original sentences with the modified and polished ones. Load the translated text processed by professional technicians and the original text into the original-translation text database.

6. A method for translating professional terminology in the ship field using an LLM fine-tuned based on qLoRA, characterized in that, The formula of the translation matching algorithm is: ； Among them, is the matching rate, is the domain matching influence coefficient, is the matching rate of the sub-library of specialized vocabulary, is the number of sub-domains, is the third consideration factor, is the first consideration factor, is the second consideration factor, is the text coherence influence coefficient; is the historical matching factor, is the historical matching influence coefficient.

7. A method for translating professional terminology in the ship field using an LLM fine-tuned based on qLoRA, characterized in that The S33 includes: S331: Extract the professional vocabulary in the text that can reflect the text theme. S332: Analyze the professional vocabulary reflecting the text theme to determine the text theme, and screen out all the original-translation comparison copies with matching text themes in the original-translation text database. S333: Conduct sentence pattern matching among all the original-translation comparison copies with matching text themes screened out, and match the original-translation comparison copy with the highest sentence pattern overlap. S334: Based on the translated text, match similar sentence patterns, find the corresponding professional vocabulary in the translated text, and obtain the historical matching factor.

8. A system applied to a method for translating professional term vocabulary in the ship field based on qLoRA fine-tuning of LLM according to any one of claims 1-7, including: Screening and calibration unit: Obtain the text to be translated, convert the text to be translated into an SQL statement, perform matching analysis with the ship field database, and screen and calibrate the professional term texts in the ship field. Professional vocabulary matching unit: Based on the professional data word library in the ship field, match the corresponding professional vocabulary groups for the professional terms calibrated in the text to be translated. Matching rate calculation unit: Through the set translation matching algorithm, calculate the matching rate of the professional vocabulary groups based on the sentence patterns where the professional vocabulary in the text to be translated is located, and screen out the vocabulary with the highest matching rate. Full text translation unit: Perform full text translation on the text to be translated, load the vocabulary with the highest matching rate into the text with partial translation completed, and complete the translation of all texts.

9. A system according to claim 8, characterized in that, The matching rate calculation unit includes: Field matching factor analysis module: Based on the screened professional term texts, analyze the sub-field of the text in the ship industry, match it with the sub-database of the ship industry, match the corresponding sub-database, match the selected vocabulary in the professional vocabulary group with the sub-database, obtain the field matching factor, and use the field matching factor as the calculation parameter of the translation matching algorithm. Text coherence factor analysis module: Load the vocabulary in the professional vocabulary group into the original-translation comparison text respectively. Based on the translated text, analyze the coherence of the text meaning through the sentence meaning translation algorithm to obtain the coherence factor of the text. Historical matching factor analysis module: According to the sub-field of the text in the ship industry, screen out the original-translation comparison texts with high matching rates in the original-translation text database. Based on the translated text, match similar sentence patterns, find the corresponding professional vocabulary in the translated text, and obtain the historical matching factor.

10. A system according to claim 8, wherein The text coherence factor analysis module includes: The first consideration factor analysis module: identifies the basic components of the translated text, analyzes the disorder rate of the sentence components after translation, and uses the disorder rate parameter of the sentence components as the first consideration factor; The second consideration factor analysis module: identifies and judges the word order of the translated text, judges the disorder rate of the word order of the sentence after translation, and uses the disorder rate of the word order as the second consideration factor; The third consideration factor analysis module: performs reverse translation on the translated text to obtain a reverse translation text, calculates the dissimilarity rate between the reverse translation text and the text to be translated, and uses the dissimilarity rate of the text to be translated as the third consideration factor; The coherence factor calculation module: calculates the final text coherence factor through a coherence algorithm based on the third consideration factor, the second consideration factor, and the first consideration factor.

Images