A bilingual corpus alignment method, system, terminal and medium
By calculating the word root hit rate between the original and translated sentences and combining it with the sentence order, the system automatically selects and aligns sentences, solving the problem of misalignment between the original and translated sentences within a paragraph and improving the accuracy and efficiency of alignment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU YOUYI INFORMATION TECH CO LTD
- Filing Date
- 2022-11-25
- Publication Date
- 2026-05-08
AI Technical Summary
In existing technologies, mismatches are easily made when the original and translated sentences within a paragraph are aligned in order, requiring manual verification, which is time-consuming, labor-intensive, and inefficient.
By calculating the word root hit rate between the original and translated sentences, setting multiple threshold groups, the sentence pair with the highest hit rate is automatically selected and output. The sentence pair is then aligned with the original sentence order. HanLP and Stanford lexical analysis are used to remove interfering words, extract word roots, and perform reverse translation to improve the alignment accuracy.
It improves the accuracy and efficiency of sentence alignment between the original text and the translation, reduces the need for manual verification, and enhances the level of automation in alignment.
Smart Images
Figure CN115796195B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of translation technology, and more specifically, to a bilingual corpus alignment method, system, terminal, and medium. Background Technology
[0002] Corpus alignment refers to matching source and translation texts at different granularities to form standardized word pairs. The units of corpus alignment range from large to small, including chapters, paragraphs, sentences, and words. The smaller the granularity of the parallel corpus, the richer the linguistic information it provides and the greater its application value. Generally, if the corpus is aligned by chapter or paragraph, the source and translation texts can be aligned sequentially. However, aligning the source and translation texts within a paragraph by sentence or even smaller granularity cannot be handled so simply. Due to various reasons such as source language style, target language style, translation style, and content adjustments, simply aligning source and translation sentences sequentially within a paragraph often results in numerous mismatches, requiring manual verification, which is time-consuming, labor-intensive, and inefficient. Summary of the Invention
[0003] The purpose of this invention is to provide a bilingual corpus alignment method, system, terminal, and medium. The method calculates the first hit rate between the effective word roots in the source sentence and the effective word roots in each translated sentence. When the first hit rate meets a first threshold group, the source sentence and translated sentence with the highest first hit rate are output according to the source sentence number. When the first hit rate meets a second threshold group, the translated sentence with the highest first hit rate is selected, and the second hit rate between the effective word roots of the translated sentence and the word roots of each source sentence is calculated. When the second hit rate meets a third threshold group, the source sentence and translated sentence with the highest second hit rate are output according to the source sentence number. When the second hit rate meets a fourth threshold group, the translated sentence is reverse-translated, and the translated words in the translated sentence are output along with the translated sentence and the source sentence with the highest word hit rate in each source sentence, according to the source sentence number. The goal is to improve the alignment accuracy of original and translated sentences by combining the hit rate between the original and translated sentences with the order of the original sentences, thereby increasing the alignment efficiency.
[0004] The above-mentioned technical objective of the present invention is achieved through the following technical solution:
[0005] A bilingual corpus alignment method includes the following steps: obtaining the sequence number and original sentence information of each original sentence in the original sentence list, and the translated sentence information of each translated sentence in the translated sentence list; wherein, the original sentence information includes a list of translated word roots of effective words in the original sentence, and the translated sentence information includes a list of word roots of effective words in the translated sentence; obtaining the first hit rate between the translated word roots in the original sentence translated word root list and the word roots in each translated sentence word root list; if the first hit rate meets a first threshold group, then outputting the original sentence and translated sentence with the highest first hit rate according to the sequence number of the original sentence; if the first hit rate meets a second threshold group, then selecting the translated sentence with the highest first hit rate. The sentence is the target object, and the second hit rate of the root words in the target object's root word list and the translated root words in the root word list of each original sentence is obtained. If the second hit rate meets the third threshold group, the original sentence with the highest second hit rate and the target object are output according to the original sentence number. If the second hit rate meets the fourth threshold group, the word list of the target object is obtained, and the words in the word list are reverse-translated according to the preset query interface to obtain the translated word list of the target object. The third hit rate of the translated words in the translated word list of the target object and the words in each original sentence is obtained, and the original sentence with the highest third hit rate and the target object are output according to the original sentence number.
[0006] Furthermore, the process of obtaining the original sentence list and the translated sentence list includes: parsing the file to be aligned to obtain the parsed original text and translated text; performing sentence segmentation on the original text and translated text respectively to obtain the corresponding original sentence list and translated sentence list; the original sentence list contains the original sentences and their corresponding serial numbers; the translated sentence list contains the translated sentences.
[0007] Furthermore, the process of obtaining the original sentence information includes: parsing the original sentence using HanLP lexical analysis and removing interference words to obtain a list of original sentence words containing valid words from the original sentence; querying the translated word list for each valid word in the original sentence word list using a preset query interface to obtain a set of translated word lists corresponding to the original sentence word list; extracting the translated word roots from the translated word lists in the translated word list set using the Stemmer algorithm to obtain a list of translated word roots corresponding to the original sentence word list; the original sentence information includes the original sentence, the original sentence word list, the set of translated word lists, and the list of translated word roots.
[0008] Furthermore, the preset query interface is prototype + root word + singular / plural.
[0009] Furthermore, the process of obtaining the translated sentence list includes: parsing the translated sentences using the Stanford lexical method and removing interference words to obtain a translated sentence word list containing the effective words of the translated sentences; extracting the root words of the effective words in the translated sentence word list using the Stemmer algorithm to obtain a root word list corresponding to the translated sentence word list; the translated sentence information includes the translated sentence, the translated sentence word list, and the root word list.
[0010] Furthermore, the method for selecting the highest hit rate includes: obtaining the highest hit rate through binary tree sorting.
[0011] A bilingual corpus alignment system, comprising:
[0012] The first acquisition module is used to acquire the sequence number and original sentence information of each original sentence in the original sentence list, and the translated sentence information of each translated sentence in the translated sentence list; wherein, the original sentence information includes a list of translated word roots of effective words in the original sentence, and the translated sentence information includes a list of word roots of effective words in the translated sentence; the second acquisition module is used to acquire the first hit rate between the translated word roots in the original sentence translated word root list and the word roots in each translated sentence word root list; the first output module is used to output the original sentence and the translated sentence with the highest first hit rate according to the sequence number of the original sentence when the first hit rate meets the first threshold group; the third acquisition module is used to select the translated sentence with the highest first hit rate when the first hit rate meets the second threshold group. The first output module is used to obtain the target object and the second hit rate of the word roots in the target object's word root list and the translated word roots in the translated word root list of each original sentence. The second output module is used to output the original sentence with the highest second hit rate and the target object according to the original sentence number when the second hit rate meets the third threshold group. The third output module is used to obtain the word list of the target object when the second hit rate meets the fourth threshold group, and to perform reverse translation on the words in the word list according to the preset query interface to obtain the translated word list of the target object. It also obtains the third hit rate of the translated words in the translated word list of the target object and the words in each original sentence, and outputs the original sentence with the highest third hit rate and the target object according to the original sentence number.
[0013] Furthermore, the third output module includes a first acquisition unit, a first processing unit, a second acquisition unit, a second processing unit, and an output unit; the first acquisition unit is connected to the third acquisition module and is used to acquire a word list of the target object when the second hit rate meets the fourth threshold group; the first processing unit is connected to the first acquisition unit and is used to perform reverse translation of the words in the word list based on the target object word list acquired by the first acquisition unit according to a preset query interface to obtain a translated word list of the target object; the second acquisition unit is connected to the first processing unit and the first acquisition module and is used to acquire the translated word list of the target object and each original sentence; the second processing unit is connected to the second acquisition unit and is used to calculate the third hit rate between the translated words in the target object's translated word list and the words in each original sentence; the output unit is connected to the second processing unit and is used to output the original sentence with the highest third hit rate and the target object according to the original sentence number.
[0014] An electronic terminal includes: a memory for storing a computer program; and a processor for executing the computer program stored in the memory to cause the electronic terminal to perform the bilingual corpus alignment method.
[0015] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the bilingual corpus alignment method.
[0016] Compared with the prior art, the present invention has the following beneficial effects:
[0017] The purpose of this invention is to provide a bilingual corpus alignment method. This method calculates the first hit rate between effective word roots in the source sentence and effective word roots in each translated sentence. When the first hit rate meets a first threshold group, the source sentence and translated sentence with the highest first hit rate are output according to the source sentence number. When the first hit rate meets a second threshold group, the translated sentence with the highest first hit rate is selected, and the second hit rate between the effective word roots of this translated sentence and the word roots of each source sentence is calculated. When the second hit rate meets a third threshold group, the source sentence and translated sentence with the highest second hit rate are output according to the source sentence number. When the second hit rate meets a fourth threshold group, the translated sentence is reverse-translated, and the translated words in the translated sentence are output along with the translated sentence and the source sentence with the highest word hit rate in each source sentence, according to the source sentence number. This achieves the goal of improving the alignment accuracy of source and translated sentences by combining the hit rate between the source and translated sentences with the order of the source sentences, thereby improving alignment efficiency. Attached Figure Description
[0018] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0019] Figure 1 This is a schematic diagram of a bilingual corpus alignment method in this embodiment;
[0020] Figure 2 This is a schematic diagram of a bilingual corpus alignment system in this embodiment.
[0021] The attached diagram shows the markings and corresponding component names:
[0022] 100 - First acquisition module; 110 - Second acquisition module; 120 - First output module; 130 - Third acquisition module; 140 - Second output module; 141 - First acquisition unit; 142 - First processing unit; 143 - Second acquisition unit; 144 - Second processing unit; 145 - Output unit. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0024] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0025] Example 1
[0026] A bilingual corpus alignment method, such as Figure 1 As shown, it includes the following steps:
[0027] S1. Obtain the sequence number and information of each original sentence in the original sentence list, and the information of each translated sentence in the translated sentence list; wherein, the original sentence information includes a list of translated word roots of the effective words in the original sentence, and the translated sentence information includes a list of translated word roots of the effective words in the translated sentence;
[0028] In this embodiment, the process of obtaining the original sentence list and the translated sentence list in S1 includes:
[0029] S11. Parse the file to be aligned to obtain the parsed original text and translated text;
[0030] S12. The original text and the translated text are processed into sentences to obtain corresponding lists of original sentences and lists of translated sentences; the list of original sentences contains the original sentences and their corresponding serial numbers; the list of translated sentences contains the translated sentences.
[0031] Specifically, the aligned file is parsed using a file parsing system to obtain the parsed original text and translated text. A text segmenter is then used to segment the original text and translated text into sentences, resulting in a list of original sentences in the original text and a list of translated sentences in the translated text. Each original sentence in the list of original sentences is assigned a sequence number according to its order in the original text.
[0032] In this embodiment, the process of obtaining the original sentence information in S1 includes:
[0033] A. Use hanlp lexical analysis to parse the original sentence and remove distracting words to obtain a list of words in the original sentence that contain the valid words of the original sentence;
[0034] Specifically, the original sentence is parsed using the hanlp lexical method to obtain a list of basic words corresponding to the original sentence. Then, punctuation marks, word combinations, and other interfering words are removed from the list of basic words to obtain a list of effective words in the original sentence.
[0035] B. Using the preset query interface, query the translation word list for each valid word in the original sentence word list to obtain the translation word list set corresponding to the original sentence word list;
[0036] Specifically, using a query interface based on prototype + root word + singular / plural, the system retrieves a list of translated terms for each valid word in the original sentence's word list from an existing Chinese-English dictionary database. This list is then compiled into a translation word set. The translation word set includes one or more translated terms for each valid word.
[0037] C. Extract the translation roots from the translation word list in the translation word list set using the Stemmer algorithm to obtain the translation root list corresponding to the original sentence word list; the original sentence information includes the original sentence, the original sentence word list, the translation word list set, and the translation root list.
[0038] For example, in the original sentence “Air China has a team of pilots and flight attendants with excellent professional skills, rigorous work style, and good service,” the effective word “business” is processed as “Word(origin=business,basic=[profess,busi,oper,professional work],target=[profession,business,professional work,operation])”, and the effective word “pilot” is processed as “Word(origin=pilot,basic=[flier,pilot,airman,aviat,airmen],target=[flier,pilots,aviators,fliers,pilot,airman,aviator,airmen])”, where origin is the effective word in the original sentence, target is the list of words in the translation, and basic is the list of word roots in the translation.
[0039] In this embodiment, the process of obtaining the list of translated sentences in S1 includes:
[0040] E. Use Stanford lexical analysis to analyze the translated sentence and remove distracting words to obtain a word list of the translated sentence containing the effective words of the translated sentence;
[0041] Specifically, the Stanford lexical method is used to analyze the translated sentence to obtain a list of basic words corresponding to the translated sentence. Interference words such as punctuation and prepositions in the list of basic words are removed to obtain a list of words in the translated sentence containing the effective words.
[0042] F. Extract the root words of valid words in the word list of the translated sentence using the Stemmer algorithm to obtain a root word list corresponding to the word list of the translated sentence; the translated sentence information includes the translated sentence, the word list of the translated sentence, and the root word list.
[0043] For example, in the sentence "Air China has a team of pilots and flight attendants with excellent business skills, precise work methods, and good service delivery," "Air" is treated as a Word(origin=Air, basic=[air], target=null), and "pilots" is treated as a Word(origin=pilots, basic=[pilot], target=null). "origin" represents the effective words in the translated sentence, "target" is the list of translated words, and "basic" is the word root.
[0044] S2. Obtain the first hit rate of the translated word roots in the original sentence translation word root list and the word roots in each translated sentence word root list;
[0045] Specifically, the translation root of each effective word in a source sentence is checked to see if it matches in the root of each translated sentence. If it matches, the word match count of the source sentence in the translated sentence is incremented by 1. The number of matched words is set as matchedWordNum, and the effective words in the source sentence are set as sourceWordnum. The first hit rate between the source sentence and each translated sentence is obtained by using the formula matchedRate = matchedWordNum / sourceWordTotal. In this way, the first hit rate between each source sentence and each translated sentence is obtained.
[0046] S3. If the first hit rate meets the first threshold group, then output the original sentence and the translated sentence with the highest first hit rate according to the original sentence number;
[0047] Specifically, if the matchedRate of the original sentence and a translated sentence is greater than or equal to 0.8 (first threshold group), then the original sentence and the translated sentence are considered to be a complete match, and the original sentence and the translated sentence are considered to be aligned bilingual corpus. The original sentence and the translated sentence are output according to the sequence number of the original sentence.
[0048] If the matchedRate of the original sentence and multiple translated sentences is greater than or equal to 0.8 (first threshold group), then the translated sentence with the highest hit rate is obtained by sorting by binary tree as the best translation, and the original sentence and translated sentence are output according to the corresponding index of the original sentence.
[0049] S4. If the first hit rate meets the second threshold group, then select the translated sentence with the highest first hit rate as the target object, and obtain the second hit rate of the word roots in the target object's word root list and the translated word roots in the word root list of each original sentence;
[0050] Specifically, if the `matchedRate` of the source sentence and all translated sentences does not meet the first threshold group, then source sentences and translated sentences with a first hit rate of 0.8 > `matchedRate` >= 0.5 (second threshold group) are considered as preliminary hits. The translated sentence with the highest first hit rate is then obtained from the preliminary hit list by sorting it using a binary tree and designated as the target object. The root words of the target object's translated sentence are queried to see if they match in the root word list of each source sentence. If they do, the number of effective words from the translated sentence that match in the source sentence is incremented by 1, and the total number of effective translated sentences is `targetWordTotal`. The second hit rate of the translated sentence with each source sentence is calculated using the formula `matchedWordNum / targetWordTotal = matchedRate`. Sentences with a first hit rate less than 0.5 are directly excluded.
[0051] S5. If the second hit rate meets the third threshold group, then output the original sentence with the highest second hit rate and the target object according to the original sentence number;
[0052] Specifically, if the matchedRate of the translated sentence and a source sentence is greater than or equal to 0.7 (third threshold group), then the translated sentence is considered to be a complete match with the source sentence. In this case, the translated sentence and the source sentence are considered to be aligned bilingual corpus, and the source sentence and the translated sentence are output according to the sequence number of the source sentence.
[0053] If the matchedRate of the translated sentence and multiple original sentences is greater than or equal to 0.7 (third threshold group), then the original sentence with the second highest hit rate is obtained by sorting by binary tree as the best original sentence, and the original sentence and the translated sentence are output according to the original sentence's index.
[0054] S6. If the second hit rate meets the fourth threshold group, then obtain the word list of the target object, and perform reverse translation on the words in the word list according to the preset query interface to obtain the translated word list of the target object;
[0055] S7. Obtain the third hit rate of the translated words in the target object's translation word list and the words in each original sentence, and output the original sentence with the highest third hit rate and the target object according to the original sentence number.
[0056] Specifically, if the matchedRate of the translated sentence and the original sentence is less than 0.5, it is discarded. If the hit rate is 0.5 <= matchedRate < 0.7 (fourth threshold group), then according to the query interface of prototype + root + singular and plural, the corresponding Chinese translation of each word in the translated sentence is obtained from the English-Chinese dictionary and matched with the Chinese words in the original sentence. The number of Chinese words in the translated sentence is compared with the number of effective Chinese words in the original sentence to obtain the third hit rate. The original sentence with the highest third hit rate and the translated sentence in that row are selected as the aligned bilingual corpus. The original sentence and the translated sentence are output according to the sequence number of the original sentence.
[0057] Example 2
[0058] A bilingual corpus alignment system, such as Figure 2 As shown, it includes: a first acquisition module 100, used to acquire the sequence number and original sentence information of each original sentence in the original sentence list, and the translated sentence information of each translated sentence in the translated sentence list; wherein, the original sentence information includes a list of translated word roots of effective words in the original sentence, and the translated sentence information includes a list of word roots of effective words in the translated sentence; a second acquisition module 110, used to acquire the first hit rate between the translated word roots in the original sentence translated word root list and the word roots in each translated sentence word root list; a first output module 120, used to output the original sentence and the translated sentence with the highest first hit rate according to the sequence number of the original sentence when the first hit rate meets the first threshold group; and a third acquisition module 130, used to select the first hit when the first hit rate meets the second threshold group. The sentence with the highest hit rate is the target object, and the second hit rate of the root words in the target object's root word list and the root words in the translated root word list of each original sentence is obtained. The second output module 140 is used to output the original sentence with the highest second hit rate and the target object according to the original sentence number when the second hit rate meets the third threshold group. The third output module is used to obtain the word list of the target object when the second hit rate meets the fourth threshold group, and to perform reverse translation on the words in the word list according to the preset query interface to obtain the translated word list of the target object. It also obtains the third hit rate of the translated words in the translated word list of the target object and the words in each original sentence, and outputs the original sentence with the highest third hit rate and the target object according to the original sentence number.
[0059] In this embodiment, the third output module includes a first acquisition unit 141, a first processing unit 142, a second acquisition unit 143, a second processing unit 144, and an output unit 145. The first acquisition unit 141 is connected to the third acquisition module 130 and is used to acquire a word list of the target object when the second hit rate meets the fourth threshold group. The first processing unit 142 is connected to the first acquisition unit 141 and is used to perform reverse translation on the words in the word list based on the target object word list acquired by the first acquisition unit 141, according to a preset query interface, to obtain a translated word list of the target object. The second acquisition unit 143 is connected to the first processing unit 142 and the first acquisition module 100 and is used to acquire the translated word list of the target object and each original sentence. The second processing unit 144 is connected to the second acquisition unit 143 and is used to calculate the third hit rate between the translated words in the target object's translated word list and the words in each original sentence. The output unit 145 is connected to the second processing unit 144 and is used to output the original sentence with the highest third hit rate and the target object according to the original sentence number.
[0060] This embodiment also provides an electronic terminal, including: a memory for storing computer programs; and a processor for executing the computer programs stored in the memory, so that the electronic terminal executes the bilingual corpus alignment method described above.
[0061] This embodiment also provides a computer-readable storage medium storing a computer program: when executed by a processor, the program implements the bilingual corpus alignment method described above.
[0062] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A bilingual corpus alignment method, characterized in that, Includes the following steps: Retrieve the sequence number and information of each original sentence in the original sentence list, and the information of each translated sentence in the translated sentence list; the original sentence information includes a list of translation root words for the effective words in the original sentence, and the translated sentence information includes a list of translation root words for the effective words in the translated sentence. Obtain the first hit rate of the translated word roots in the original sentence translation word root list and the word roots in each translated sentence word root list; If the first hit rate meets the first threshold group, then output the original sentence and the translated sentence with the highest first hit rate according to the original sentence number; If the first hit rate meets the second threshold group, then the translated sentence with the highest first hit rate is selected as the target object, and the second hit rate of the word roots in the target object's word root list and the translated word roots in the word root list of each original sentence is obtained; If the second hit rate meets the third threshold group, then output the original sentence with the highest second hit rate and the target object according to the original sentence number; If the second hit rate meets the fourth threshold group, then obtain the word list of the target object, and perform reverse translation on the words in the word list according to the preset query interface to obtain the translated word list of the target object. Obtain the third hit rate of the translated words in the target object's translated word list and the words in each original sentence, and output the original sentence with the highest third hit rate and the target object according to the original sentence number.
2. The bilingual corpus alignment method according to claim 1, characterized in that, The process of obtaining the list of original sentences and the list of translated sentences includes: The file to be aligned is parsed to obtain the parsed original text and translated text. The original text and the translated text are processed into sentences separately to obtain the corresponding lists of original sentences and translated sentences; The list of original sentences includes the original sentences and their corresponding serial numbers; The list of translated sentences contains the translated sentences.
3. The bilingual corpus alignment method according to claim 1, characterized in that, The process of obtaining information from the original sentence includes: The original sentence is parsed using the hanlp lexical method, and interference words are removed to obtain a list of words in the original sentence that contain the effective words of the original sentence; Using the preset query interface, query the translation word list for each valid word in the original sentence word list to obtain the translation word list set corresponding to the original sentence word list; The Stemmer algorithm is used to extract the root words of the translated words in the translated word list set to obtain a list of root words corresponding to the original sentence word list; The original sentence information includes the original sentence, a list of words in the original sentence, a list of words in the translated sentence, and a list of word roots in the translated sentence.
4. A bilingual corpus alignment method according to any one of claims 1 or 3, characterized in that: The preset query interface is prototype + root word + singular / plural.
5. The bilingual corpus alignment method according to claim 1, characterized in that, The process of obtaining the list of translated sentences includes: The Stanford lexical analysis was used to analyze the translated sentence and remove distracting words to obtain a list of words in the translated sentence that contains the effective words of the translated sentence; The Stemmer algorithm is used to extract the root words of valid words in the word list of the translated sentence to obtain a root word list corresponding to the word list of the translated sentence. The translated sentence information includes the translated sentence, a list of words in the translated sentence, and a list of word roots.
6. The bilingual corpus alignment method according to claim 1, characterized in that, The selection method with the highest hit rate includes: The highest hit rate is obtained by sorting using a binary tree.
7. A bilingual corpus alignment system, characterized in that, include: The first acquisition module (100) is used to acquire the serial number and original sentence information of each original sentence in the original sentence list, and the translated sentence information of each translated sentence in the translated sentence list; wherein, the original sentence information includes the translation root list of the effective words of the original sentence, and the translated sentence information includes the word root list of the effective words of the translated sentence; The second acquisition module (110) is used to acquire the first hit rate of the translated word roots in the original sentence translation word root list and the word roots in each translated sentence word root list; The first output module (120) is used to output the original sentence and the translated sentence with the highest hit rate according to the original sentence number when the first hit rate meets the first threshold group. The third acquisition module (130) is used to select the translated sentence with the highest first hit rate as the target object when the first hit rate meets the second threshold group, and to obtain the second hit rate of the word roots in the word root list of the target object and the translated word roots in the word root list of each original sentence; The second output module (140) is used to output the original sentence with the highest second hit rate and the target object according to the original sentence number when the second hit rate meets the third threshold group. The third output module is used to obtain the word list of the target object when the second hit rate meets the fourth threshold group, and to perform reverse translation on the words in the word list according to the preset query interface to obtain the translated word list of the target object. It also obtains the third hit rate of the translated words in the target object's translated word list and the words in each original sentence, and outputs the original sentence with the highest third hit rate and the target object according to the original sentence number.
8. A bilingual corpus alignment system according to claim 7, characterized in that: The third output module includes a first acquisition unit (141), a first processing unit (142), a second acquisition unit (143), a second processing unit (144), and an output unit (145); The first acquisition unit (141) is connected to the third acquisition module (130) and is used to acquire the word list of the target object when the second hit rate meets the fourth threshold group; The first processing unit (142) is connected to the first acquisition unit (141) and is used to perform reverse translation of the words in the word list based on the target object word list acquired by the first acquisition unit (141) according to a preset query interface, so as to obtain the translated word list of the target object; The second acquisition unit (143) is connected to the first processing unit (142) and the first acquisition module (100), and is used to acquire the translation word list of the target object and each original sentence; The second processing unit (144) is connected to the second acquisition unit (143) and is used to calculate the third hit rate between the translated words in the target object translation word list and the words in each original sentence; The output unit (145) is connected to the second processing unit (144) and is used to output the third original sentence with the highest hit rate and the target object according to the original sentence number.
9. An electronic terminal, characterized in that, include: The memory is used to store computer programs; A processor for executing a computer program stored in the memory to cause an electronic terminal to perform a bilingual corpus alignment method according to any one of claims 1-6.
10. A computer-readable storage medium storing a computer program thereon, characterized in that: When the program is executed by the processor, it implements a bilingual corpus alignment method according to any one of claims 1-6.
Citation Information
Patent Citations
Domain self-adaption sentence alignment system based on self-guidance mode
CN103678565A
Sentence alignment method for bilingual parallel corpuses
CN107391495A