English long text translation segmentation strategy optimization method and system based on deep learning
Through the deep learning-based English long text translation segmentation strategy optimization system, the problems of inefficiency and poor quality of traditional translation methods when processing long texts are solved, and the translation efficiency and quality improvement is achieved.
Patent Information
- Application Number
- CN202510085057.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-20
- Publication Date
- 2025-05-30
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional translation methods are inefficient and of uneven quality when processing long English texts, which affects translation progress and accuracy.
A deep learning-based English long text translation segmentation strategy optimization system is adopted. This system automatically determines and adjusts the text segmentation length through the collaboration of determining modules, translation modules, judgment modules and adjustment modules to improve translation efficiency and quality.
Through intelligent optimization of segmentation strategies, the system can improve the quality and efficiency of translation and ensure the accuracy and reliability of translation results.
Smart Images
Figure CN120068891A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of strategy optimization, and more particularly, to a method and system for optimizing the segmentation strategy of English long text translation based on deep learning. Background Art
[0002] With the continuous advancement of globalization, the demand for English long text translation has shown a significant growth trend. However, in the actual operation process, traditional translation methods often encounter difficult problems such as low translation efficiency and uneven translation quality when dealing with long text. These problems not only affect the progress of translation work but also have a negative impact on the accuracy and reliability of translation.
[0003] Therefore, it is necessary to design a method and system for optimizing the segmentation strategy of English long text translation based on deep learning to solve the problems existing in the current technology. Summary of the Invention
[0004] In view of this, the present invention proposes a method and system for optimizing the segmentation strategy of English long text translation based on deep learning, aiming to improve translation efficiency.
[0005] On the one hand, the present invention proposes a system for optimizing the segmentation strategy of English long text translation based on deep learning, including:
[0006] A determination module, configured to determine an English long text to be translated, randomly extract a sample to be translated from the English long text to be translated, collect text feature data of the sample to be translated, and determine an initial text segmentation length of the English long text to be translated according to the text feature data;
[0007] A translation module, configured to translate the sample to be translated based on a deep learning algorithm to obtain a translation quality score;
[0008] A judgment module, configured to judge whether to adjust the initial text segmentation length according to the translation quality score;
[0009] An adjustment module, configured to collect a translation duration for translating the sample to be translated when the judgment module determines to adjust the initial text segmentation length, calculate a comprehensive score value according to the translation duration and the translation quality score, determine an adjustment coefficient of the initial text segmentation length according to the comprehensive score value; and adjust the initial text segmentation length according to the adjustment coefficient.
[0010] Furthermore, the text feature data includes: lexical feature data, syntactic feature data;
[0011] The lexical feature data includes: lexical length feature data, word frequency data, lexical richness data, stop word ratio data;
[0012] The syntactic feature data includes: sentence length feature data, sentence type feature data.
[0013] Further, when determining the initial text segmentation length of the long English text to be translated according to the text feature data, it includes:
[0014] Performing feature extraction on the text feature data to obtain a lexical length feature value, a word frequency feature value, a lexical richness feature value, a stop word ratio feature value, a sentence length feature value, and a sentence type feature value;
[0015] Performing weighted summation on the lexical length feature value, the word frequency feature value, the lexical richness feature value, the stop word ratio feature value, the sentence length feature value, and the sentence type feature value to obtain a comprehensive text feature value;
[0016] Determining the initial text segmentation length of the long English text to be translated according to the text feature value.
[0017] Further, when determining the initial text segmentation length of the long English text to be translated according to the text feature value, it includes:
[0018] Comparing the text feature value with a first text feature value and a second text feature value, and determining the initial text segmentation length according to the comparison result; wherein, the first text feature value is less than the second text feature value;
[0019] When the text feature value is less than or equal to the first text feature value, determining the initial text segmentation length as a first segmentation length;
[0020] When the text feature value is greater than the first text feature value and less than or equal to the second text feature value, determining the initial text segmentation length as a second segmentation length;
[0021] When the text feature value is greater than the second text feature value, determining the initial text segmentation length as a third segmentation length.
[0022] Further, when determining whether to adjust the initial text segmentation length according to the translation quality score, it includes:
[0023] The translation quality score is obtained by the following formula:
[0024] P = ω1×A + ω2×L + ω3×G + ω4×D;
[0025] Among them, P represents the translation quality score; A represents the accuracy score; L represents the fluency score; G represents the grammar consistency score; D represents the vocabulary diversity score; ω1 represents the first weight coefficient; ω2 represents the second weight coefficient; ω3 represents the third weight coefficient; ω4 represents the fourth weight coefficient.
[0026] Further, when determining whether to adjust the segmentation length of the initial text according to the translation quality score, it further includes:
[0027] Comparing the translation quality score with a translation quality score threshold, and determining whether to adjust the segmentation length of the initial text according to the comparison result;
[0028] When the translation quality score is greater than or equal to the translation quality score threshold, it is determined not to adjust the segmentation length of the initial text;
[0029] When the translation quality score is less than the translation quality score threshold, it is determined to adjust the segmentation length of the initial text.
[0030] Further, when determining the adjustment coefficient of the segmentation length of the initial text based on the comprehensive score value, it includes:
[0031] The comprehensive score value is obtained by the following formula:
[0032]
[0033] Among them, M represents the comprehensive score value; P represents the translation quality score; T represents the translation duration.
[0034] Further, when determining the adjustment coefficient of the segmentation length of the initial text based on the comprehensive score value, it further includes:
[0035] Comparing the comprehensive score value with historical data, and determining the adjustment coefficient of the segmentation length of the initial text according to the comparison result;
[0036] When there is a historical comprehensive score value in the historical data that is the same as the comprehensive score value, adjusting the segmentation length of the initial text according to the historical adjustment coefficient corresponding to the historical comprehensive score value, and obtaining the text segmentation adjustment length;
[0037] When there is no historical comprehensive score value in the historical data that is the same as the comprehensive score value, determining the adjustment coefficient according to the comprehensive score value, and adjusting the segmentation length of the initial text according to the adjustment coefficient to obtain the text segmentation adjustment length.
[0038] Further, when determining the adjustment coefficient according to the comprehensive score value, it includes:
[0039] Compare the comprehensive score value with the first comprehensive score value and the second comprehensive score value, and determine the adjustment coefficient according to the comparison result; wherein, the first comprehensive score value is less than the second comprehensive score value.
[0040] When the comprehensive score value is less than or equal to the first comprehensive score value, determine the adjustment coefficient as the first adjustment coefficient, and take the product value of the first adjustment coefficient and the initial text segmentation length as the text segmentation adjustment length.
[0041] When the comprehensive score value is greater than the first comprehensive score value and less than or equal to the second comprehensive score value, determine the adjustment coefficient as the second adjustment coefficient, and take the product value of the second adjustment coefficient and the initial text segmentation length as the text segmentation adjustment length.
[0042] When the comprehensive score value is greater than the second comprehensive score value, determine the adjustment coefficient as the third adjustment coefficient, and take the product value of the third adjustment coefficient and the initial text segmentation length as the text segmentation adjustment length.
[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: The system for optimizing the English long text translation segmentation strategy based on deep learning provided by the present invention realizes the intelligent optimization of the English long text translation segmentation strategy through the mutual cooperation of the determination module, the translation module, the judgment module and the adjustment module. In the determination module, the system can automatically identify and extract samples of the English long text to be translated. By collecting and analyzing the text feature data of these samples, the system can initially determine a reasonable text segmentation length. This makes full use of deep learning technology, which can automatically learn and identify the impact of different text features on the translation effect, so as to make a more accurate segmentation decision. The translation module is based on deep learning algorithms to perform high-quality translation on the samples to be translated and give a translation quality score. This score not only considers the accuracy of the translation, but also comprehensively considers multiple dimensions such as the fluency of the translation, grammar consistency and vocabulary diversity, so as to more comprehensively evaluate the quality of the translation. The judgment module judges whether it is necessary to adjust the initial text segmentation length according to the translation quality score. If the translation quality score is high, it means that the current segmentation strategy is already relatively reasonable and no excessive adjustment is required; if the translation quality score is low, it means that there may be problems with the current segmentation strategy and adjustment is needed to optimize the translation effect. In the adjustment module, the system will calculate a comprehensive score value according to the translation duration and the translation quality score, and determine the adjustment coefficient according to this comprehensive score value. The size of the adjustment coefficient will directly affect the adjustment range of the initial text segmentation length. Through this step, the system can automatically adjust the segmentation strategy to adapt to the impact of different text features on the translation effect, thereby improving the quality and efficiency of the translation.
[0044] In another aspect, the present invention also proposes an optimization method for the segmentation strategy of long English text translation based on deep learning, including the following steps:
[0045] S100: Determine the long English text to be translated, randomly extract the translation samples of the long English text to be translated, collect the text feature data of the translation samples, and determine the initial text segmentation length of the long English text to be translated according to the text feature data;
[0046] S200: Translate the translation samples based on the deep learning algorithm to obtain a translation quality score;
[0047] S300: Judge whether to adjust the initial text segmentation length according to the translation quality score;
[0048] S400: When the judgment module determines to adjust the initial text segmentation length, collect the translation duration of translating the translation samples, calculate a comprehensive score value according to the translation duration and the translation quality score, and determine the adjustment coefficient of the initial text segmentation length according to the comprehensive score value;
[0049] S500: Adjust the initial text segmentation length according to the adjustment coefficient.
[0050] It can be understood that the above optimization method and system for the segmentation strategy of long English text translation based on deep learning have the same beneficial effects, which will not be elaborated here. Description of the Drawings
[0051] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered as limiting the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:
[0052] Figure 1 is the structural block diagram of the optimization system for the segmentation strategy of long English text translation based on deep learning provided by the embodiment of the present invention;
[0053] Figure 2 is the flowchart of the optimization method for the segmentation strategy of long English text translation based on deep learning provided by the embodiment of the present invention. Detailed Embodiments
[0054] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the drawings and in conjunction with the embodiments.
[0055] Referring to Figure 1 As shown, in some embodiments of the present application, this embodiment provides an optimization system for the segmentation strategy of long English text translation based on deep learning, including:
[0056] A determination module, configured to determine a long English text to be translated, randomly extract a sample to be translated from the long English text to be translated, collect text feature data of the sample to be translated, and determine an initial text segmentation length of the long English text to be translated according to the text feature data;
[0057] A translation module, configured to translate the sample to be translated based on a deep learning algorithm and obtain a translation quality score;
[0058] A judgment module, configured to judge whether to adjust the initial text segmentation length according to the translation quality score;
[0059] An adjustment module, configured to, when the judgment module determines to adjust the initial text segmentation length, collect a translation duration for translating the sample to be translated, calculate a comprehensive score value according to the translation duration and the translation quality score, determine an adjustment coefficient of the initial text segmentation length according to the comprehensive score value; and adjust the initial text segmentation length according to the adjustment coefficient.
[0060] It can be understood that the optimization system for the English long text translation segmentation strategy provided in this embodiment realizes the intelligent optimization of the English long text translation segmentation strategy through the mutual cooperation of the determination module, the translation module, the judgment module, and the adjustment module. In the determination module, the system can automatically identify and extract samples of the English long text to be translated. By collecting and analyzing the text feature data of these samples, the system can initially determine a reasonable text segmentation length. This makes full use of deep learning technology, which can automatically learn and identify the impact of different text features on the translation effect, thereby making a more accurate segmentation decision. The translation module, based on deep learning algorithms, performs high-quality translation on the samples to be translated and gives a translation quality score. This score not only considers the accuracy of the translation but also comprehensively considers multiple dimensions such as the fluency of the translation, grammatical consistency, and lexical diversity, so as to more comprehensively evaluate the quality of the translation. The judgment module then determines whether it is necessary to adjust the initial text segmentation length according to the translation quality score. If the translation quality score is high, it indicates that the current segmentation strategy is already relatively reasonable and does not require excessive adjustment; if the translation quality score is low, it indicates that there may be problems with the current segmentation strategy and adjustments need to be made to optimize the translation effect. In the adjustment module, the system calculates a comprehensive score value based on the translation duration and the translation quality score, and determines the adjustment coefficient according to this comprehensive score value. The magnitude of the adjustment coefficient will directly affect the adjustment range of the initial text segmentation length. Through this step, the system can automatically adjust the segmentation strategy to adapt to the impact of different text features on the translation effect, thereby improving the quality and efficiency of the translation.
[0061] Specifically, the text feature data includes: lexical feature data, grammatical feature data;
[0062] The lexical feature data includes: lexical length feature data, word frequency data, lexical richness data, stop word ratio data;
[0063] The grammatical feature data includes: sentence length feature data, sentence type feature data.
[0064] Specifically, when determining the initial text segmentation length of the English long text to be translated according to the text feature data, it includes:
[0065] Performing feature extraction on the text feature data to obtain lexical length feature values, word frequency feature values, lexical richness feature values, stop word ratio feature values, sentence length feature values, and sentence type feature values;
[0066] Performing weighted summation on the lexical length feature values, word frequency feature values, lexical richness feature values, stop word ratio feature values, sentence length feature values, and sentence type feature values to obtain a comprehensive text feature value;
[0067] Determine the initial text segmentation length of the long English text to be translated according to the text feature values.
[0068] It can be understood that the lexical length feature value is preferably the average value or median of the lexical lengths, used to characterize the average length or length distribution of the words in the long English text to be translated; the word frequency feature value is preferably the maximum frequency value of the words appearing in the long English text to be translated, used to characterize the usage frequency of the words in the long English text to be translated; the lexical richness feature value is preferably the ratio of the number of different words to the total number of words in the long English text to be translated, used to characterize the diversity degree of the words in the long English text to be translated; the stop word ratio feature value is preferably the ratio of the number of stop words to the total number of words in the long English text to be translated, used to characterize the proportion of useless words in the long English text to be translated; the sentence length feature value is preferably the average value or median of the sentence lengths, used to characterize the average length or length distribution of the sentences in the long English text to be translated; the sentence type feature value is preferably the ratio of the number of different sentence types to the total number of sentences in the long English text to be translated, used to characterize the diversity degree of the sentence types in the long English text to be translated. By comprehensively considering these text feature values, the system can more accurately determine the initial text segmentation length of the long English text to be translated, providing a more reliable basis for subsequent optimization and adjustment.
[0069] Specifically, when determining the initial text segmentation length of the long English text to be translated according to the text feature values, it includes:
[0070] Compare the text feature values with a first text feature value and a second text feature value, and determine the initial text segmentation length according to the comparison result; wherein, the first text feature value is less than the second text feature value.
[0071] When the text feature value is less than or equal to the first text feature value, determine the initial text segmentation length as the first segmentation length.
[0072] When the text feature value is greater than the first text feature value and less than or equal to the second text feature value, determine the initial text segmentation length as the second segmentation length.
[0073] When the text feature value is greater than the second text feature value, determine the initial text segmentation length as the third segmentation length.
[0074] It can be understood that the first segment length, the second segment length, and the third segment length are preset according to different ranges of text feature values, and they represent different text segmentation strategies. For example, for a text with a lower text feature value, it may mean that the text is relatively simple and the sentence structure is relatively single. Therefore, a shorter segment length, that is, the first segment length, can be selected to facilitate the translation module to perform fast and accurate translation. For a text with a higher text feature value, it may mean that the text is relatively complex and the sentence structure is relatively diverse. Therefore, a longer segment length, that is, the third segment length, can be selected to reduce the problem of loss of context information caused by excessive segmentation. When the text feature value is between the two, the second segment length is selected as the initial text segment length. Through such a segmentation strategy, the system can more flexibly adapt to text translation tasks of different difficulties and improve the accuracy and efficiency of translation.
[0075] Specifically, when judging whether to adjust the initial text segment length according to the translation quality score, it includes:
[0076] The translation quality score is obtained by the following formula:
[0077] P = ω1×A + ω2×L + ω3×G + ω4×D;
[0078] Wherein, P represents the translation quality score; A represents the accuracy score; L represents the fluency score; G represents the grammar consistency score; D represents the vocabulary diversity score; ω1 represents the first weight coefficient; ω2 represents the second weight coefficient; ω3 represents the third weight coefficient; ω4 represents the fourth weight coefficient.
[0079] It can be understood that in the calculation of the translation quality score, each dimension (accuracy, fluency, grammar consistency, and vocabulary diversity) is assigned corresponding weight coefficients (ω1, ω2, ω3, ω4), and these weight coefficients reflect the importance of different dimensions in evaluating the translation quality. By multiplying the scores of these dimensions by the corresponding weight coefficients and summing them, a comprehensive translation quality score can be obtained. This score can comprehensively reflect the quality of the translation and provide a reliable basis for subsequent adjustment steps.
[0080] In this embodiment, the accuracy score (A) is measured using the BLEU score or other similar metrics commonly used in machine translation evaluation, which reflects the matching degree between the translation result and the reference translation. The fluency score (L) is obtained by evaluating the smoothness and readability of the translated text, ensuring that the translated text is natural and fluent in grammar and expression. The grammar consistency score (G) focuses on the correctness of the grammatical structure of the translated text, ensuring that there are no obvious grammar errors in the translation result. The vocabulary diversity score (D) is evaluated by calculating the proportion of different words used in the translated text, which encourages the use of a more diverse vocabulary in the translation result to improve the expressive effect and readability of the translation.
[0081] In this embodiment, the preferred value of ω1 is 0.4, the preferred value of ω2 is 0.2, the preferred value of ω3 is 0.2, and the preferred value of ω4 is 0.2. Such a weight assignment takes into account both the accuracy of the translation and also takes into account the fluency, grammar consistency, and vocabulary diversity of the translation, and can more comprehensively evaluate the quality of the translation. Of course, according to different application scenarios and translation requirements, these weight coefficients can also be adjusted accordingly to adapt to different evaluation criteria. By adopting such a translation quality scoring calculation method, the system can more accurately judge the translation quality under the current segmentation strategy, thereby providing a more precise basis for subsequent adjustment steps.
[0082] Specifically, when determining whether to adjust the initial text segmentation length according to the translation quality score, it further includes:
[0083] Comparing the translation quality score with a translation quality score threshold, and determining whether to adjust the initial text segmentation length according to the comparison result;
[0084] When the translation quality score is greater than or equal to the translation quality score threshold, it is determined not to adjust the initial text segmentation length;
[0085] When the translation quality score is less than the translation quality score threshold, it is determined to adjust the initial text segmentation length.
[0086] It can be understood that the translation quality score threshold is obtained based on a large amount of translation practice experience and data statistical analysis, and it represents an acceptable translation quality level. When the translation quality score is higher than or equal to this threshold, it indicates that the current segmentation strategy can already produce translation results of relatively high quality and no further adjustment is required. When the translation quality score is lower than this threshold, it indicates that there may be problems with the current segmentation strategy, resulting in poor translation quality, so adjustment is needed to optimize the translation effect. Through such a judgment mechanism, the system can automatically adjust the segmentation strategy according to the translation quality score to ensure the accuracy and efficiency of the translation result.
[0087] Specifically, when determining the adjustment coefficient of the initial text segmentation length based on the comprehensive score value, it includes:
[0088] The comprehensive score value is obtained by the following formula:
[0089]
[0090] Where M represents the comprehensive score value; P represents the translation quality score; T represents the translation duration.
[0091] It can be seen that the calculation method of the comprehensive score value M takes into account both the translation quality score P and introduces the factor of translation duration T. By multiplying the translation quality score by the translation duration and multiplying by a coefficient (here it is 1000, used to adjust the value range for easy understanding and application), the system can obtain a comprehensive score value that reflects both translation quality and translation efficiency. This comprehensive score value can provide a more comprehensive basis for determining the adjustment coefficient subsequently.
[0092] Specifically, when determining the adjustment coefficient of the initial text segmentation length based on the comprehensive score value, it further includes:
[0093] Compare the comprehensive score value with historical data, and determine the adjustment coefficient of the initial text segmentation length according to the comparison result;
[0094] When there is a historical comprehensive score value in the historical data that is the same as the comprehensive score value, adjust the initial text segmentation length according to the historical adjustment coefficient corresponding to the historical comprehensive score value, and obtain the text segmentation adjustment length;
[0095] When there is no historical comprehensive score value in the historical data that is the same as the comprehensive score value, determine the adjustment coefficient according to the comprehensive score value, and adjust the initial text segmentation length according to the adjustment coefficient to obtain the text segmentation adjustment length.
[0096] It is understandable that by comparing with historical data, the system can utilize past experience and data to guide current adjustment decisions. If there are historical comprehensive score values in the historical data that are the same as or similar to the current comprehensive score value, the system can directly adopt the corresponding historical adjustment coefficient to adjust the initial text segment length, thereby quickly and accurately determining the text segment adjustment length. Such a method not only improves the efficiency of adjustment but also ensures that the adjustment effect is consistent with past experience, reducing the uncertainty brought by subjective judgment. When there is no historical comprehensive score value in the historical data that is the same as the current comprehensive score value, the system will calculate the adjustment coefficient based on the current comprehensive score value. This calculation process may involve complex mathematical models or algorithms to ensure that the adjustment coefficient can accurately reflect the impact of the comprehensive score value on the adjustment of the initial text segment length. By adopting such a method, the system can more flexibly adapt to different translation tasks and text characteristics, ensuring the accuracy and efficiency of the translation results.
[0097] In this embodiment, the historical data includes information such as historical comprehensive score values, historical adjustment coefficients, and historical text feature data. These information are stored and updated by the system for quick retrieval and application in subsequent translation tasks.
[0098] Specifically, when determining the adjustment coefficient according to the comprehensive score value, it includes:
[0099] Compare the comprehensive score value with a first comprehensive score value and a second comprehensive score value, and determine the adjustment coefficient according to the comparison result; wherein, the first comprehensive score value is less than the second comprehensive score value;
[0100] When the comprehensive score value is less than or equal to the first comprehensive score value, determine the adjustment coefficient as the first adjustment coefficient, and take the product value of the first adjustment coefficient and the initial text segment length as the text segment adjustment length;
[0101] When the comprehensive score value is greater than the first comprehensive score value and less than or equal to the second comprehensive score value, determine the adjustment coefficient as the second adjustment coefficient, and take the product value of the second adjustment coefficient and the initial text segment length as the text segment adjustment length;
[0102] When the comprehensive score value is greater than the second comprehensive score value, determine the adjustment coefficient as the third adjustment coefficient, and take the product value of the third adjustment coefficient and the initial text segment length as the text segment adjustment length.
[0103] It can be understood that the first adjustment coefficient, the second adjustment coefficient, and the third adjustment coefficient are preset according to different comprehensive score value ranges, and they represent different adjustment amplitudes. For example, for a text with a relatively low comprehensive score value, it may indicate low translation quality or low translation efficiency. Therefore, a relatively large adjustment coefficient, that is, the first adjustment coefficient, can be selected to make a relatively large adjustment to the initial text segment length, so as to quickly optimize the translation effect. For a text with a relatively high comprehensive score value, it may indicate relatively high translation quality or good translation efficiency. Therefore, a relatively small adjustment coefficient, that is, the third adjustment coefficient, can be selected to avoid excessive adjustment to the initial text segment length and maintain the stability and consistency of the translation. When the comprehensive score value is between the two, the second adjustment coefficient is selected as the adjustment coefficient to make a moderate adjustment to the initial text segment length. Through such an adjustment strategy, the system can more precisely control the adjustment amplitude, ensuring the accuracy and efficiency of the translation result. At the same time, the implementation of this segmentation strategy and optimization method has also greatly improved the quality and efficiency of English long text translation, providing strong support for cross-language communication and text processing.
[0104] Refer to Figure 2 As shown, in some embodiments of the present application, this embodiment provides an optimization method for a segmentation strategy of English long text translation based on deep learning, including the following steps:
[0105] S100: Determine the English long text to be translated, randomly extract the translation samples of the English long text to be translated, collect the text feature data of the translation samples, and determine the initial text segment length of the English long text to be translated according to the text feature data;
[0106] S200: Translate the translation samples based on a deep learning algorithm to obtain a translation quality score;
[0107] S300: Determine whether to adjust the initial text segment length according to the translation quality score;
[0108] S400: When the judgment module determines that the initial text segment length needs to be adjusted, collect the translation duration of translating the translation samples, calculate a comprehensive score value according to the translation duration and the translation quality score, and determine the adjustment coefficient of the initial text segment length according to the comprehensive score value;
[0109] S500: Adjust the initial text segment length according to the adjustment coefficient.
[0110] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.
[0111] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0112] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0113] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the specific implementation manners of the present invention can still be modified or equivalently replaced, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.
Claims
1. A deep learning-based English long text translation segmentation strategy optimization system, characterized in that: include: A determination module is configured to determine an English long text to be translated, randomly extract a sample to be translated of the English long text to be translated, collect text feature data of the sample to be translated, and determine an initial text segment length of the English long text to be translated according to the text feature data; A translation module is configured to translate the sample to be translated based on a deep learning algorithm to obtain a translation quality score; A judgment module, configured to judge whether to adjust the length of the initial text segment according to the translation quality score; The adjustment module is configured to collect the translation time of the sample to be translated when the judgment module determines to adjust the length of the initial text segment, calculate a comprehensive score value according to the translation time and the translation quality score, determine the adjustment coefficient of the length of the initial text segment according to the comprehensive score; and adjust the length of the initial text segment according to the adjustment coefficient.
2. The English long text translation segmentation strategy optimization system based on deep learning according to claim 1 is characterized in that: The text feature data includes: vocabulary feature data and grammar feature data; The vocabulary feature data includes: vocabulary length feature data, word frequency data, vocabulary richness data, and stop word ratio data; The grammatical feature data includes: sentence length feature data and sentence type feature data.
3. The English long text translation segmentation strategy optimization system based on deep learning according to claim 2 is characterized in that: When determining the initial text segment length of the long English text to be translated according to the text feature data, it includes: Performing feature extraction on the text feature data to obtain a vocabulary length feature value, a word frequency feature value, a vocabulary richness feature value, a stop word ratio feature value, a sentence length feature value, and a sentence type feature value; Performing weighted summation on the vocabulary length feature value, word frequency feature value, vocabulary richness feature value, stop word ratio feature value, sentence length feature value, and sentence type feature value to obtain a comprehensive text feature value; The initial text segment length of the long English text to be translated is determined according to the text feature value.
4. The English long text translation segmentation strategy optimization system based on deep learning according to claim 3 is characterized in that: When determining the initial text segment length of the long English text to be translated according to the text feature value, it includes: Comparing the text feature value with the first text feature value and the second text feature value, and determining the initial text segment length according to the comparison result; wherein the first text feature value is smaller than the second text feature value; When the text feature value is less than or equal to the first text feature value, determining the initial text segment length to be the first segment length; When the text feature value is greater than the first text feature value and less than or equal to the second text feature value, determining the initial text segment length to be the second segment length; When the text feature value is greater than the second text feature value, the initial text segment length is determined to be a third segment length.
5. The English long text translation segmentation strategy optimization system based on deep learning according to claim 1 is characterized in that: When determining whether to adjust the length of the initial text segment according to the translation quality score, it includes: The translation quality score is obtained by the following formula: P=ω1×A+ω2×L+ω3×G+ω4×D; Among them, P represents the translation quality score; A represents the accuracy score; L represents the fluency score; G represents the grammatical consistency score; D represents the lexical diversity score; ω1 represents the first weight coefficient; ω2 represents the second weight coefficient; ω3 represents the third weight coefficient; ω4 represents the fourth weight coefficient.
6. The English long text translation segmentation strategy optimization system based on deep learning according to claim 5 is characterized in that: When determining whether to adjust the length of the initial text segment according to the translation quality score, the method further includes: Comparing the translation quality score with a translation quality score threshold, and determining whether to adjust the length of the initial text segment according to the comparison result; When the translation quality score is greater than or equal to the translation quality score threshold, determining not to adjust the initial text segment length; When the translation quality score is less than the translation quality score threshold, it is determined to adjust the length of the initial text segment.
7. The English long text translation segmentation strategy optimization system based on deep learning according to claim 6 is characterized in that: When determining the adjustment coefficient of the initial text segment length based on the comprehensive score value, it includes: The comprehensive score is obtained by the following formula: Among them, M represents the comprehensive score; P represents the translation quality score; T represents the translation time.
8. The English long text translation segmentation strategy optimization system based on deep learning according to claim 6 is characterized in that: When determining the adjustment coefficient of the initial text segment length based on the comprehensive score value, it also includes: Comparing the comprehensive score value with historical data, and determining the adjustment coefficient of the initial text segment length according to the comparison result; When there is a historical comprehensive score value identical to the comprehensive score value in the historical data, adjusting the initial text segment length according to the historical adjustment coefficient corresponding to the historical comprehensive score value, and obtaining the text segment adjustment length; When there is no historical comprehensive score value identical to the comprehensive score value in the historical data, the adjustment coefficient is determined according to the comprehensive score value pair, and the initial text segment length is adjusted according to the adjustment coefficient to obtain the text segment adjusted length.
9. The English long text translation segmentation strategy optimization system based on deep learning according to claim 8 is characterized in that: When determining the adjustment coefficient according to the comprehensive score value pair, it includes: Comparing the comprehensive score value with a first comprehensive score value and a second comprehensive score value, and determining the adjustment coefficient according to the comparison result; wherein the first comprehensive score value is less than the second comprehensive score value; When the comprehensive score value is less than or equal to the first comprehensive score value, determining the adjustment coefficient to be the first adjustment coefficient, and taking the product of the first adjustment coefficient and the initial text segment length as the text segment adjustment length; When the comprehensive score value is greater than the first comprehensive score value and less than or equal to the second comprehensive score value, determining the adjustment coefficient to be the second adjustment coefficient, and taking the product of the second adjustment coefficient and the initial text segment length as the text segment adjustment length; When the comprehensive score value is greater than the second comprehensive score value, the adjustment coefficient is determined to be a third adjustment coefficient, and a product value of the third adjustment coefficient and the initial text segment length is used as the text segment adjustment length.
10. A method for optimizing segmentation strategy of English long text translation based on deep learning, applied to the system for optimizing segmentation strategy of English long text translation based on deep learning as claimed in any one of claims 1 to 9, characterized in that: include: Determine an English long text to be translated, randomly extract a sample to be translated of the English long text to be translated, collect text feature data of the sample to be translated, and determine the initial text segment length of the English long text to be translated according to the text feature data; Translate the sample to be translated based on a deep learning algorithm to obtain a translation quality score; Determining whether to adjust the length of the initial text segment according to the translation quality score; When the judgment module determines to adjust the length of the initial text segment, the translation time of the sample to be translated is collected, and a comprehensive score is calculated according to the translation time and the translation quality score, and the adjustment coefficient of the initial text segment length is determined according to the comprehensive score; The initial text segment length is adjusted according to the adjustment coefficient.