A translation scoring method and related device

Abstract

The present application discloses a translation scoring method and related equipment, the method comprising: after obtaining the user translation text input by the terminal user for the text to be translated, firstly pre-matching the user translation text with at least one reference translation text of the text to be translated, and obtaining the matching representation data corresponding to each reference translation text; then, according to the matching representation data corresponding to each reference translation text, respectively determine the matching score corresponding to each reference translation text; finally, according to the matching score corresponding to at least one reference translation text, determine the translation score of the user translation text. Among them, because at least one matching score can more accurately represent the matching degree between the user translation text and the diversified translation results of the text to be translated, the translation score determined based on these matching scores can more accurately represent the matching degree between the user translation text and the text to be translated, which is conducive to improving the accuracy of the translation score.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a translation scoring method and related equipment. Background Art

[0002] With the popularity of online language education (for example, English, Japanese, etc.), more and more students tend to choose to learn and assess languages ​​online. Online translation scoring not only saves teachers a lot of time and energy, but also overcomes the defect that different teachers give inconsistent scores for the same answer due to different scoring standards, which is conducive to improving the fairness of language education. This makes online translation scoring a focus of attention in the field of education.

[0003] However, due to defects in the translation scoring technology, the translation score determined based on the translation scoring technology is inaccurate. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a translation scoring method and related devices, which can improve the accuracy of translation scoring.

[0005] The present application provides a translation scoring method, the method comprising:

[0006] After obtaining a user translation text input by a terminal user for the text to be translated, performing a preset matching process on the user translation text and at least one reference translation text of the text to be translated, respectively, to obtain matching representation data corresponding to each reference translation text;

[0007] Determining the matching scores of the reference translation texts according to the matching representation data corresponding to the reference translation texts;

[0008] A translation score of the user translated text is determined according to the matching score corresponding to the at least one reference translated text.

[0009] In a possible implementation, the at least one reference translation text includes a target translation text, and a process of determining matching representation data corresponding to the target translation text includes:

[0010] Performing word matching processing on the user translated text and the target translated text to obtain a word matching result corresponding to the target translated text;

[0011] Performing word matching processing on the user translated text and the target translated text to obtain a word matching result corresponding to the target translated text;

[0012] The character matching result and the word matching result are combined to obtain matching representation data corresponding to the target translation text.

[0013] In a possible implementation, the process of determining the word matching result includes:

[0014] Preprocessing the user translated text to obtain a word segmentation set of the user translated text;

[0015] Preprocessing the target translation text to obtain a word set of the target translation text;

[0016] Performing a first word matching process on the word set of the user translation text and the word set of the target translation text to obtain a first word matching result;

[0017] Performing a second word matching process on the word set of the user translation text and the word set of the target translation text to obtain a second word matching result;

[0018] The first word segmentation matching result and the second word segmentation matching result are combined to obtain a word matching result corresponding to the target translation text.

[0019] In a possible implementation, the preprocessing includes word segmentation processing and filtering processing.

[0020] In a possible implementation, the at least one reference translation text includes a target translation text, and a process of determining a matching score corresponding to the target translation text includes:

[0021] Determining the translation type of the text to be translated;

[0022] Searching for a scoring model to be used corresponding to the translation type from the preset mapping relationship; wherein the preset mapping relationship includes a correspondence between the translation type and the pre-constructed scoring model to be used;

[0023] Using the scoring model to be used, scoring the matching representation data corresponding to the target translation text to obtain a model scoring result corresponding to the target translation text;

[0024] According to the model scoring result, a matching score corresponding to the target translation text is determined.

[0025] In a possible implementation, the matching representation data corresponding to the target translation text includes first representation data and second representation data; wherein the first representation data is used to represent a word matching result between the user translation text and the target translation text; and the second representation data is used to represent a word order matching result between the user translation text and the target translation text;

[0026] The process of determining the model scoring result includes:

[0027] If it is determined that the translation type of the text to be translated is a sentence translation type, inputting the first representation data into the scoring model to be used to obtain a model scoring result corresponding to the target translation text output by the scoring model to be used;

[0028] If it is determined that the translation type of the text to be translated is a short text translation type, the matching representation data corresponding to the target translation text is input into the scoring model to be used to obtain a model scoring result corresponding to the target translation text output by the scoring model to be used.

[0029] In a possible implementation, the process of determining the matching score includes:

[0030] Determining a correction parameter value corresponding to the target translated text according to a similarity between the information quantity characterization data of the user translated text and the information quantity characterization data of the target translated text;

[0031] The matching score corresponding to the target translation text is determined according to the product of the model scoring result and the correction parameter value.

[0032] In a possible implementation manner, the information quantity characterization data includes text length and word count, and the process of determining the correction parameter value includes:

[0033] Determining a text length ratio according to a ratio between a text length of the user translated text and a text length of the target translated text;

[0034] Determining a vocabulary ratio according to a ratio between the number of vocabulary in the user translation text and the number of vocabulary in the target translation text;

[0035] The correction parameter value corresponding to the target translation text is determined according to the sum of the text length ratio and the vocabulary number ratio.

[0036] In a possible implementation manner, the matching representation data corresponding to the target translation text includes second representation data; wherein the second representation data is used to represent a word order matching result between the user translation text and the target translation text;

[0037] The process of determining the matching score includes:

[0038] If it is determined that the second representation data is lower than the first representation threshold, determining a matching score corresponding to the target translation text according to the model scoring result and a first preset scoring value;

[0039] If it is determined that the second representation data is not lower than the first representation threshold and lower than the second representation threshold, determining a matching score corresponding to the target translation text according to the model scoring result and a second preset scoring value;

[0040] If it is determined that the second representation data is not lower than a second representation threshold, a matching score corresponding to the target translation text is determined according to the model scoring result.

[0041] In a possible implementation, the process of determining the translation score includes:

[0042] Performing preset data processing on the matching score corresponding to the at least one reference translation text to obtain a score to be used;

[0043] According to the preset full score of the text to be translated, the score to be used is converted into a score system to obtain the translation score of the user translated text.

[0044] In a possible implementation, the process of generating the at least one reference translation text includes:

[0045] Obtaining a standard translation text of the text to be translated;

[0046] Performing word segmentation processing on the standard translation text to obtain a vocabulary set of the standard translation text;

[0047] Searching for a synonym for at least one word in the vocabulary from a pre-built synonym database;

[0048] At least one reference translation text of the text to be translated is generated according to the standard translation text and the synonym of the at least one vocabulary.

[0049] The present application also provides a translation scoring device, including:

[0050] A content matching unit, configured to, after acquiring a user translation text input by a terminal user for a text to be translated, perform preset matching processing on the user translation text and at least one reference translation text of the text to be translated, respectively, to obtain matching representation data corresponding to each reference translation text;

[0051] A translation scoring unit, configured to determine a matching score corresponding to each of the reference translation texts according to the matching representation data corresponding to each of the reference translation texts;

[0052] The score determination unit is used to determine the translation score of the user translated text according to the matching score corresponding to the at least one reference translated text.

[0053] The embodiment of the present application further provides a device, the device comprising: a processor, a memory, and a system bus;

[0054] The processor and the memory are connected via the system bus;

[0055] The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes any implementation of the translation scoring method provided in the embodiments of the present application.

[0056] The embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any implementation of the translation scoring method provided in the embodiment of the present application.

[0057] The present application also provides a computer program product. When the computer program product is run on a terminal device, the terminal device executes any implementation of the translation scoring method provided in the present application.

[0058] Based on the above technical solution, this application has the following beneficial effects:

[0059] In the technical solution provided by the present application, after obtaining the user translation text input by the terminal user for the text to be translated, the user translation text is firstly matched with at least one reference translation text of the text to be translated respectively to obtain the matching representation data corresponding to each reference translation text; then, the matching score corresponding to each reference translation text is determined respectively according to the matching representation data corresponding to each reference translation text; finally, the translation score of the user translation text is determined according to the matching score corresponding to at least one reference translation text. Among them, because the above-mentioned "at least one reference translation text" can accurately represent the diversified translation results of the text to be translated, the matching score determined based on the "at least one reference translation text" can more accurately represent the matching degree between the user translation text and the diversified translation results of the text to be translated, so that the translation score determined based on these matching scores can more accurately represent the matching degree between the user translation text and the text to be translated, which is conducive to improving the accuracy of the translation score. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 A flowchart of a translation scoring method provided in an embodiment of the present application;

[0062] Figure 2 A schematic diagram of a translation scoring process provided in an embodiment of the present application;

[0063] Figure 3 A schematic diagram of the structure of a translation scoring device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The inventor found in the research of the above-mentioned "translation scoring technology" that for the above-mentioned "translation scoring technology", after obtaining the user translation result of the student's answer to a certain source language translation test question (such as an English translation test question), the translation score of the user translation result can usually be determined based on the difference between the user translation result and the unique standard answer set in advance for the source language translation test question. However, for some language translation (for example, English to Chinese, Japanese to Chinese) scenarios, due to the diversity of expressions in the target language (for example, Chinese) (for example, multiple words express the same meaning, etc.), a certain source language translation test question may correspond to multiple target language translation results, so that the above-mentioned "unique standard answer" cannot represent the above-mentioned "multiple target language translation results", so that the translation score determined based on the relevant translation scoring technology cannot accurately represent the difference between the user translation result and the above-mentioned "multiple target language translation results", and thus the translation score determined based on the relevant translation scoring technology is inaccurate.

[0065] Based on the above findings, in order to solve the technical problems in the background technology part, the embodiment of the present application provides a translation scoring method, which includes: after obtaining the user translation text input by the terminal user for the text to be translated, first pre-matching the user translation text with at least one reference translation text of the text to be translated, and obtaining the matching representation data corresponding to each reference translation text; then, according to the matching representation data corresponding to each reference translation text, respectively determine the matching score corresponding to each reference translation text; finally, according to the matching score corresponding to at least one reference translation text, determine the translation score of the user translation text. Among them, because the above-mentioned "at least one reference translation text" can accurately represent the diversified translation results of the text to be translated, the matching score determined based on the "at least one reference translation text" can more accurately represent the matching degree between the user translation text and the diversified translation results of the text to be translated, so that the translation score determined based on these matching scores can more accurately represent the matching degree between the user translation text and the text to be translated, which is conducive to improving the accuracy of the translation score.

[0066] In addition, the embodiments of the present application do not limit the execution subject of the translation scoring method. For example, the translation scoring method provided in the embodiments of the present application can be applied to data processing devices such as terminal devices or servers. Among them, the terminal device can be a smart phone, a computer, a personal digital assistant (PDA) or a tablet computer. The server can be an independent server, a cluster server or a cloud server.

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0068] Method Embodiment

[0069] See also Figure 1 , which is a flow chart of a translation scoring method provided in an embodiment of the present application.

[0070] The translation scoring method provided in the embodiment of the present application includes S1-S3:

[0071] S1: After obtaining a user translation text input by a terminal user for a text to be translated, performing preset matching processing on the user translation text and at least one reference translation text of the text to be translated to obtain matching representation data corresponding to each reference translation text.

[0072] In S1, "end user" refers to the user of the language translation test device (for example, any student who is learning a language online). "Language translation test device" refers to the terminal device used for language translation test (for example, a smart phone, computer, tablet computer or learning machine, etc.). "Language translation" refers to translating text data in a source language (for example, English) into text data in a target language (for example, Chinese).

[0073] In addition, the embodiment of the present application does not limit the relationship between the above-mentioned "language translation test device" and the execution subject of the translation scoring method. For example, in some application scenarios, the "language translation test device" can be used as the execution subject of the translation scoring method, that is, the "language translation test device" can execute any implementation of the translation scoring method provided in the embodiment of the present application. For example, in other application scenarios, the server can be used as the execution subject of the translation scoring method. At this time, the "language translation test device" can communicate with the server so that the "language translation test device" can send the above-mentioned "text to be translated" and "user translation text" to the server, so that the server can first determine the translation score of the user translation text by executing any implementation of the translation scoring method provided in the embodiment of the present application, and then feed back the translation score of the user translation text to the "language translation test device" so that the "language translation test device" sends the translation score of the user translation text to the terminal user in a preset manner (for example, screen display, SMS notification, email notification, etc.).

[0074] When the server executes any implementation of the translation scoring method provided in the embodiment of the present application, the "text to be translated" in S1 refers to the source language text data that needs to be translated by the terminal user when the terminal user performs a source language translation test (e.g., an English translation test) (e.g., an English test text that needs to be answered by the terminal user). In addition, the embodiment of the present application does not limit the "text to be translated", for example, it may include at least one sentence.

[0075] In S1, the "user translated text" refers to the target language text data obtained by translating the above "text to be translated" by the terminal user (that is, the translation answer provided by the terminal user for the above "text to be translated"). In addition, the embodiment of the present application does not limit the input method of the "user translated text", and can be implemented by any existing or future input method (such as text input method, language input method or picture input method, etc.).

[0076] The "reference translation text" in S1 refers to the actual translation result of the above-mentioned "text to be translated" in the target language (that is, the reference answer of the "text to be translated"). In addition, the embodiment of the present application does not limit the generation method of "at least one reference translation text", for example, it can be pre-set. For another example, in order to improve the flexibility of generating reference answers, it can specifically include steps 11-12:

[0077] Step 11: Obtain the standard translation text of the text to be translated.

[0078] Among them, "standard translation text" refers to the standard answer set in advance for the text to be translated.

[0079] Step 12: Perform word segmentation processing on the standard translation text of the text to be translated to obtain a vocabulary set of the standard translation text.

[0080] The "vocabulary set" is used to record the vocabulary in the standard translation text.

[0081] In addition, the embodiment of the present application is not limited to the "word segmentation processing" in step 12, and can be implemented using any existing or future word segmentation method (for example, stuttering word segmentation, etc.).

[0082] Step 13: Find synonyms of at least one word in the vocabulary set from a pre-built synonym database.

[0083] The "synonym database" is used to record synonyms of at least one word; and the embodiment of the present application does not limit the representation method of the "synonym database", for example, it can be represented in the form of an array. That is, in the "synonym database", a word (for example, "love") and all synonyms of the word (for example, "love" and "cherish") can be put into the same array, so that all words in each array (for example, "love", "love" and "cherish") are synonyms of each other.

[0084] In addition, the embodiment of the present application does not limit the implementation method of step 13. For example, when the above-mentioned "vocabulary set" includes the first word to the Nth word, step 13 may specifically include steps 131 to 134:

[0085] Step 131: Search for a synonym of the nth word in a synonym database to obtain a search result for the nth word, where n is a positive integer, n≤N, and N is a positive integer.

[0086] Among them, "the search result of the nth word" refers to the search result of the synonyms of the nth word; and the embodiment of the present application does not limit the specific content of the "search result of the nth word", for example, it may include whether there is a synonym of the nth word in the synonym database, the position of the synonym of the nth word in the database, and at least one of the synonyms of the nth word.

[0087] Step 132: If the search result of the nth word indicates that a synonym of the nth word exists in the synonym database, then at least one synonym of the nth word is determined according to the search result of the nth word, where n is a positive integer, n≤N, and N is a positive integer.

[0088] Based on the relevant content of the above step 13, it can be known that after obtaining the vocabulary set of the standard translation text, synonyms of each word in the vocabulary set can be searched in the synonym database to obtain synonyms of at least one word in the vocabulary set, so that the above-mentioned "at least one reference translation text" can be generated later by utilizing these synonyms and the above-mentioned "standard translation text".

[0089] Step 14: Generate at least one reference translation text of the text to be translated based on the standard translation text of the text to be translated and the above-mentioned “at least one synonym of the vocabulary”.

[0090] As an example, assume that the above “synonyms of at least one word” include M synonyms. Based on this assumption, it can be known that step 14 may specifically include steps 141 and 142:

[0091] Step 141: Generate H replacement word sets based on M synonyms, so that each replacement word set includes at least one synonym, and any two replacement word sets in the H replacement word sets are different from each other, where M is a positive integer and H is a positive integer.

[0092] Step 142: Using each replacement word in the hth replacement word set, replace the text vocabulary corresponding to each replacement word in the standard translation text to obtain the hth reference translation text, so that the hth reference translation text includes each replacement word in the hth replacement word set. Wherein, h is a positive integer, h≤H, and H is a positive integer.

[0093] Here, "replacement word" refers to the vocabulary that exists in the replacement word set.

[0094] "Text vocabulary" refers to the vocabulary that exists in the standard translation text.

[0095] The gth word in the hth replacement word set h The text vocabulary corresponding to the replacement word is the one that exists in the standard translation text and is related to the g h The replacement words are synonyms of each other.

[0096] Based on the relevant content of the above step 14, it can be known that after obtaining the synonyms corresponding to the vocabulary set (that is, the above-mentioned "synonyms of at least one vocabulary"), at least one reference translation text can be generated by replacing synonyms with the standard translation text, so that the at least one reference translation text can accurately represent the diverse translation answers of the text to be translated, which is conducive to improving the accuracy of the reference translation answers of the text to be translated.

[0097] The "preset matching process" in S1 is used to perform matching process on two text data; and the embodiment of the present application does not limit the "preset matching process", for example, it may include character matching process and / or word matching process.

[0098] The matching representation data corresponding to the qth reference translation text is used to indicate the matching degree (or similarity degree) between the user translation text and the qth reference translation text, where q is a positive integer, q≤Q, Q is a positive integer, and Q represents the number of reference translation texts.

[0099] In addition, since the matching representation data corresponding to each reference translation text is determined in a similar manner, for ease of understanding, the following description is made by taking the matching representation data corresponding to the target translation text as an example. The “target translation text” may refer to any reference translation text in the above “at least one reference translation text”.

[0100] As an example, the “process of determining the matching representation data corresponding to the target translation text” may specifically include steps 21 to 23:

[0101] Step 21: Perform word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text.

[0102] The “character matching result corresponding to the target translation text” is used to indicate the degree of matching between the text information carried by the user translation text and the text information carried by the target translation text.

[0103] In addition, the embodiment of the present application does not limit the implementation method of "word matching processing". For example, if the user translation text includes Y words, step 21 may specifically include: after obtaining the word set of the target translation text, match the yth word in the user translation text with at least one word in the word set to obtain the word matching result of the yth word; if the matching result of the yth word indicates that there is a word in the word set that successfully matches the yth word, then delete the "word that successfully matches the yth word" from the word set so that the "word that successfully matches the yth word" does not exist in the word set, and continue to perform the word matching process for the y+1th word in the user translation text using the word set; if the matching result of the yth word indicates that there is no word in the word set that successfully matches the yth word, then the word matching process for the y+1th word in the user translation text can be directly performed. Among them, the above-mentioned "word set of the target translation text" is used to record each word in the target translation text. y is a positive integer, y≤Y, and Y is a positive integer.

[0104] It can be seen that after obtaining the text set of the user translation text and the target translation text, each character in the user translation text can be matched with the characters in the text set in turn to obtain the character matching results of each character in the user translation text, so that the character matching results of these characters can comprehensively represent the degree of character matching between the user translation text and the target translation text. In addition, because the above-mentioned "text set" will be updated as the successfully matched characters appear in the user translation text, there will be no characters in the above-mentioned "text set" that have been successfully matched with at least one character in the user translation text, which can effectively avoid the adverse effects of repeated characters on the character matching process (for example, when the user translation text is "today today today", and the target translation text is "the air is so good today", if the "text set" is not deleted, it will cause each character in the user translation text to be successfully matched with the target translation text, thus causing inaccurate character matching results).

[0105] Based on the relevant content of the above step 21, it can be known that after obtaining the user translation text, each character in the user translation text can be matched with at least one character in the target translation text to obtain and output the word matching result corresponding to the target translation text, so that the word matching result can represent the degree of matching between the text information carried by the user translation text and the text information carried by the target translation text.

[0106] Step 22: Perform word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text.

[0107] The “word matching result corresponding to the target translation text” is used to indicate the matching degree between the vocabulary information carried by the user translation text and the vocabulary information carried by the target translation text.

[0108] In addition, the embodiment of the present application does not limit the implementation method of the "word matching process" in step 22, and any existing or future word matching method (for example, BLEU (bilingual evaluation under study) algorithm) can be used for implementation. For ease of understanding, an example is provided below for explanation.

[0109] As an example, step 22 may specifically include steps 221 to 225:

[0110] Step 221: pre-process the user translation text to obtain a word segmentation set of the user translation text.

[0111] The "pre-processing" can be preset, for example, it can specifically include: word segmentation processing and filtering processing. The "filtering processing" is used to filter out useless words (for example, function words, etc.) for a word segmentation set.

[0112] The “word segmentation set of the user translated text” is used to record at least one word in the user translated text.

[0113] In addition, the embodiment of the present application does not limit the implementation method of step 221. For example, it may specifically include: first performing word segmentation processing on the user translation text to obtain the word segmentation processing result of the user translation text, so that the word segmentation processing result includes each word segmentation of the user translation text; then filtering the word segmentation processing result of the user translation text to obtain a word segmentation set of the user translation text, so that there are no useless words (for example, function words) in the word segmentation set, thereby effectively avoiding the interference of useless words in the user translation text with the subsequent word matching processing process, thereby helping to improve the accuracy of word matching.

[0114] Step 222: pre-process the target translation text to obtain a word segmentation set of the target translation text.

[0115] It should be noted that the implementation of step 222 is similar to the implementation of step 221 above, and it is only necessary to replace the "user translation text" in any implementation of step 221 above with the "target translation text".

[0116] Step 223: Perform a first word matching process on the word set of the user translation text and the word set of the target translation text to obtain a first word matching result.

[0117] The "first word matching process" may be preset; for example, the "first word matching process" may be implemented using a unigram model (ie, a 1-gram algorithm).

[0118] The “first word segmentation matching result” is used to indicate the degree of vocabulary matching between the user translation text and the target translation text.

[0119] Step 224: Perform a second word matching process on the word set of the user translation text and the word set of the target translation text to obtain a second word matching result.

[0120] The "second word matching process" may be preset; for example, the "second word matching process" may be implemented using a bigram model (ie, a 2-gram algorithm).

[0121] The “second word segmentation matching result” is used to indicate the degree of matching between the vocabulary description information of the user translation text (e.g., vocabulary and its word order) and the vocabulary description information of the target translation text (e.g., used to indicate the degree of matching between all vocabulary and word order in the user translation text and all vocabulary and word order in the target translation text).

[0122] Step 225: Merge the first word segmentation matching result and the second word segmentation matching result to obtain a word matching result corresponding to the target translation text.

[0123] In an embodiment of the present application, after obtaining the first word segmentation matching result and the second word segmentation matching result, the two word matching results can be merged to obtain the word matching result corresponding to the target translation text, so that the word matching result can accurately represent the degree of matching between the vocabulary information carried by the user translation text and the vocabulary information carried by the target translation text.

[0124] Based on the relevant content of the above step 22, it can be known that after obtaining the user translation text and the target translation text, the two texts can be word matched to obtain the word matching result corresponding to the target translation text, so that the word matching result can accurately represent the degree of matching between the vocabulary information carried by the user translation text and the vocabulary information carried by the target translation text.

[0125] Step 23: Merge the character matching result corresponding to the target translation text with the word matching result corresponding to the target translation text to determine the matching representation data corresponding to the target translation text.

[0126] In an embodiment of the present application, after obtaining the character matching result corresponding to the target translation text and the word matching result corresponding to the target translation text, the character matching result and the word matching result can be merged to obtain matching representation data corresponding to the target translation text, so that the matching representation data can accurately represent the degree of matching between the user translation text and the target translation text, so that the score value achieved by the user translation text relative to the target translation text can be determined based on the matching representation data.

[0127] Based on the relevant contents of the above steps 21 to 23, it can be known that after obtaining the user translated text and the target translated text, a preset matching process (such as character matching process and word matching process) can be performed on the two text data to obtain matching representation data corresponding to the target translated text, so that the matching representation data can accurately represent the degree of matching between the user translated text and the target translated text, so that the score value achieved by the user translated text relative to the target translated text can be determined based on the matching representation data.

[0128] It should be noted that the matching representation data corresponding to any reference translation text in the embodiments of the present application are implemented using any of the execution processes shown in the above steps 21 to 23. That is, the process of determining the matching representation data corresponding to the qth reference translation text can be implemented using any of the above-mentioned implementation methods of the "process of determining the matching representation data corresponding to the target translation text", and it is only necessary to replace the "target translation text" with the "qth reference translation text" in any of the above-mentioned implementation methods of the "process of determining the matching representation data corresponding to the target translation text". Wherein, q is a positive integer, q≤Q, Q is a positive integer, and Q represents the number of reference translation texts.

[0129] Based on the relevant contents of S1 above, it can be known that when a terminal user (for example, a student) inputs a user translation text (for example, Figure 2 After that, the execution device of the "translation scoring method" may compare the user translation text with various reference translation texts of the text to be translated (for example, Figure 2 The reference answers 1, 2, ..., Q, etc. shown in the figure are respectively matched to obtain matching representation data corresponding to each reference translation text (for example, Figure 2 The “word matching degree”, “1-gram word matching degree” and “2-gram word matching degree” shown in the figure can be used to determine the translation score of the user's translated text based on these matching representation data.

[0130] S2: Determine the matching scores corresponding to the reference translation texts according to the matching representation data corresponding to the reference translation texts.

[0131] The matching score corresponding to the qth reference translation text is used to represent the score value obtained by the user translation text relative to the qth reference translation text, so that the matching score can accurately represent the similarity between the user translation text and the qth reference translation text. Where q is a positive integer, q≤Q, Q is a positive integer, and Q represents the number of reference translation texts.

[0132] In addition, there is a positive correlation between the "matching score corresponding to the qth reference translation text" and the matching representation data corresponding to the qth reference translation text, specifically: when the matching representation data includes a character matching result, a first word segmentation matching result, and a first word segmentation matching result, the "matching score corresponding to the qth reference translation text" is positively correlated with the character matching result, the first word segmentation matching result, and the first word segmentation matching result corresponding to the qth reference translation text, respectively.

[0133] In addition, the embodiment of the present application does not limit the method for determining the "matching score corresponding to the qth reference translation text", and any existing or future positive correlation algorithm may be used for implementation.

[0134] In fact, different types of language translation questions have different assessment directions. For example, single-sentence translation questions (that is, language translation questions belonging to the sentence translation type) mainly examine the use of words, but multi-sentence translation questions (that is, language translation questions belonging to the short text translation type) not only examine the use of words but also need to examine the semantic connection between different sentences. Based on this, in order to improve the accuracy of translation scoring, different translation scoring standards can be used for different types of language translation questions (for example, when grading the translation of user answers to single-sentence translation questions, the degree of character matching and the degree of word matching can be emphasized. For another example, when grading the translation of user answers to multi-sentence translation questions, the degree of character matching, the degree of word matching, and the order of arrangement between different words can be emphasized).

[0135] In order to achieve the above requirements, the embodiment of the present application also provides a possible implementation method for determining a "matching score", and for ease of understanding, the following is an example of a process for determining a matching score corresponding to a target translation text.

[0136] As an example, the “process of determining the matching score corresponding to the target translation text” may specifically include steps 31 to 34:

[0137] Step 31: Determine the translation type of the text to be translated.

[0138] Among them, the "translation type of the text to be translated" refers to the question type of the text to be translated; and the embodiment of the present application does not limit the "translation type of the text to be translated", for example, it can be a sentence translation type or a short article translation type.

[0139] In addition, the embodiment of the present application does not limit the implementation method of step 31. For ease of understanding, two examples are used for illustration below.

[0140] Example 1: If the text length of the text to be translated is greater than a preset length threshold, the short text translation type is determined as the "translation type of the text to be translated"; if the text length of the text to be translated is not greater than (that is, less than or equal to) the preset length threshold, the sentence translation type is determined as the "translation type of the text to be translated".

[0141] Example 2: If there is only one preset punctuation mark in the text to be translated (for example, punctuation marks such as “.” and “?” used to mark the end of a sentence), the sentence translation type can be determined as the “translation type of the text to be translated”; if there are at least two preset punctuation marks in the text to be translated, the short text translation type can be determined as the “translation type of the text to be translated”.

[0142] Based on the relevant content of the above step 31, it can be known that after obtaining the text to be translated, the translation type of the text to be translated can be determined based on the text length of the text to be translated and / or the punctuation marks in the text to be translated, so that the translation score of the above "user translated text" can be performed according to the translation scoring standard corresponding to the translation type.

[0143] Step 32: Find the scoring model to be used corresponding to the translation type from the preset mapping relationship. The preset mapping relationship includes the corresponding relationship between the translation type and the pre-built scoring model to be used.

[0144] Among them, the "preset mapping relationship" is used to record the translation scoring models corresponding to different translation types; and the embodiment of the present application does not limit the "preset mapping relationship", for example, it can specifically include: the correspondence between sentence translation type and sentence scoring model, and the correspondence between essay translation type and essay scoring model.

[0145] The "sentence scoring model" refers to a model that performs scoring processing according to the translation scoring standard corresponding to the sentence translation type, so that the "sentence scoring model" is suitable for performing translation scoring processing on the user translation results of a sentence translation test question.

[0146] It can be seen that the “sentence scoring model” can be used to perform scoring processing on the input data of the “sentence scoring model”; and the embodiments of the present application do not limit the working principle of the “sentence scoring model”. For example, it can be implemented using formulas (1)-(2).

[0147]

[0148]

[0149] In the formula, Indicates that the sentence scoring model is used for its input data I sentence The scoring results obtained; sentence Represents the input data of the sentence scoring model, and the I sentence It may include some of the data in the above-mentioned "matching characterization data", for example, V char Indicates the word matching result in the above-mentioned "matching representation data"; Indicates the first word matching result in the above “matching representation data”; α 1 , α 2 , α 3 and α 4 are all model parameters of the sentence scoring model; and α 4 <α 3 <α 2 <α 1 ; F base Indicates a pre-set benchmark score (e.g., F base =2).

[0150] It should be noted that the present embodiment does not limit 1 , α 2 , α 3 and α 4 , for example, for the pre-built sentence scoring model, α 1 can be 0.751, α 2 can be 0.655, α 3 can be 0.56, and α 4 It can be 0.34.

[0151] The "sentence scoring model" can be constructed in advance based on the sample sentence translation text, the actual translation text corresponding to the sample sentence translation text, and the actual scoring result of the sample sentence translation text. Among them, the "sample sentence translation text" refers to the translation response obtained by the sample user to the sample sentence translation test question. The "actual translation text corresponding to the sample sentence translation text" refers to the actual translation answer to the above-mentioned "sample sentence translation test question". The "actual scoring result of the sample sentence translation text" is used to indicate the actual score of the "sample sentence translation text" relative to the "actual translation text corresponding to the sample sentence translation text"; and the embodiment of the present application does not limit the method of obtaining the "actual scoring result of the sample sentence translation text", for example, it can be implemented by manual scoring by a teacher.

[0152] In addition, the embodiments of the present application do not limit the construction process of the "sentence scoring model". For example, it may specifically include: firstly pre-matching the sample sentence translation text with the actual translation text corresponding to the sample sentence translation text to obtain the matching representation data of the sample sentence translation text; then using the matching representation data of the sample sentence translation text and the actual scoring result of the sample sentence translation text, the sentence scoring model is trained so that the scoring result output by the trained sentence scoring model for the matching representation data of the sample sentence translation text is very close to (or even equal to) the actual scoring result of the sample sentence translation text, so that the trained sentence scoring model has better scoring performance. It can be seen that the above α 1 , α2 , α 3 and α 4 Both can be determined with the help of the training process of the sentence scoring model.

[0153] The "short article scoring model" refers to a model that performs scoring processing according to the translation scoring standard corresponding to the short article translation type, so that the "short article scoring model" is suitable for performing translation scoring processing on the user translation results of a short article translation test question.

[0154] It can be seen that the “short essay scoring model” can be used to score the input data of the “short essay scoring model”; and the embodiment of the present application does not limit the working principle of the “short essay scoring model”, for example, it can be implemented using formula (3).

[0155]

[0156] In the formula, Indicates the use of the short article scoring model for its input data I shortessay The scoring results obtained; shortessay represents the input data of the short article scoring model, and the I shortessay It can include all the data in the above-mentioned "matching characterization data", for example, V char Indicates the word matching result in the above-mentioned "matching representation data"; Indicates the first word segmentation matching result in the above “matching representation data”; represents the second word segmentation matching result in the above “matching representation data”; β 1 , β 2 , β 3 and β 4 are all model parameters of the short article scoring model.

[0157] It should be noted that the embodiments of the present application are not limited; 1 , β 2 , β 3 and β 4 For example, for the pre-built short article scoring model, β 1 Can be -0.25, β 2 can be 1.30, β 3 can be 1.30, and β 4 It can be 13.74870481.

[0158] The "short essay scoring model" can be constructed in advance based on the sample short essay translation text, the actual translation text corresponding to the sample short essay translation text, and the actual scoring result of the sample short essay translation text. Among them, the "sample short essay translation text" refers to the translation response obtained by the sample user to the sample short essay translation test question. The "actual translation text corresponding to the sample short essay translation text" refers to the actual translation answer to the above-mentioned "sample short essay translation test question". The "actual scoring result of the sample short essay translation text" is used to indicate the actual score of the "sample short essay translation text" relative to the "actual translation text corresponding to the sample short essay translation text"; and the embodiment of the present application does not limit the method of obtaining the "actual scoring result of the sample short essay translation text", for example, it can be implemented by manual scoring by teachers.

[0159] In addition, the embodiments of the present application do not limit the construction process of the "short article scoring model". For example, it may specifically include: first pre-matching the sample short article translation text with the actual translation text corresponding to the sample short article translation text to obtain the matching representation data of the sample short article translation text; then using the matching representation data of the sample short article translation text and the actual scoring result of the sample short article translation text, the short article scoring model is trained so that the translation scoring result output by the trained short article scoring model for the matching representation data of the sample short article translation text is very close to (or even equal to) the actual scoring result of the sample short article translation text, so that the trained short article scoring model has better scoring performance. It can be seen that the above β 1 , β 2 , β 3 and β 4 All of these can be determined with the help of the training process of the short article scoring model.

[0160] In step 32, the "scoring model to be used" refers to a scoring model that can be used in the translation scoring process for the user-translated text of the text to be translated. For example, if the text to be translated belongs to the sentence translation type, the scoring model to be used can be a sentence scoring model; if the text to be translated belongs to the short text translation type, the scoring model to be used can be a short text scoring model.

[0161] Based on the relevant content of the above step 32, it can be known that after obtaining the translation type of the text to be translated, the scoring model to be used corresponding to the translation type can be determined with the help of a pre-constructed preset mapping relationship, so that the scoring model to be used can perform scoring processing according to the translation scoring standard corresponding to the translation type.

[0162] Step 33: Determine the model scoring result corresponding to the target translation text according to the matching representation data corresponding to the target translation text and the scoring model to be used.

[0163] The “model scoring result” refers to the scoring result of the user translation text relative to the target translation text, so that the “model scoring result” can indicate the degree of match between the user translation text and the target translation text.

[0164] In addition, the embodiment of the present application does not limit step 33, and in order to facilitate the understanding of step 33, an example is provided below for explanation.

[0165] Assume that the “matching representation data corresponding to the target translation text” includes the first representation data and the second representation data. The “first representation data” is used to represent the word matching result between the user translation text and the target translation text. The second representation data is used to represent the word order matching result between the user translation text and the target translation text.

[0166] It should be noted that the embodiments of the present application do not limit the "first representation data" and the "second representation data". For example, the "first representation data" may include the character matching result corresponding to the target translation text and the first word segmentation matching result corresponding to the target translation text, and the "second representation data" may include the second word segmentation matching result corresponding to the target translation text.

[0167] As an example, based on the above assumptions, step 33 may specifically include steps 331 and 332:

[0168] Step 331: If it is determined that the above “translation type of the text to be translated” is a sentence translation type, the first representation data corresponding to the target translation text is input into the scoring model to be used, and a model scoring result corresponding to the target translation text output by the scoring model to be used is obtained.

[0169] It can be seen that after determining that the translation type of the text to be translated is a sentence translation type, only part of the data in the above-mentioned "matching representation data corresponding to the target translation text" (for example, the character matching result and the first word segmentation matching result) can be directly input into the scoring model to be used (that is, the sentence scoring model), so that the scoring model to be used can perform translation scoring processing on this part of the data (such as the scoring processing process shown in formulas (1)-(2)), obtain and output the model scoring result corresponding to the target translation text, so that the model scoring result can represent the score value obtained by the user translation text relative to the target translation text.

[0170] Step 332: If it is determined that the above “translation type of the text to be translated” is a short text translation type, the matching representation data corresponding to the target translation text is input into the scoring model to be used, and a model scoring result corresponding to the target translation text output by the scoring model to be used is obtained.

[0171] It can be seen that after determining that the translation type of the text to be translated is a sentence translation type, all the data in the above-mentioned "matching representation data corresponding to the target translation text" (for example, the character matching result, the first word segmentation matching result and the second word segmentation matching result) can be directly input into the scoring model to be used (that is, the short article scoring model), so that the scoring model to be used can perform translation scoring processing on this part of the data (such as the scoring processing process shown in formula (3)), obtain and output the model scoring result corresponding to the target translation text, so that the model scoring result can represent the score value obtained by the user translation text relative to the target translation text.

[0172] Based on the relevant content of the above step 33, it can be known that after obtaining the matching representation data corresponding to the target translation text, all or part of the matching representation data can be input into the scoring model to be used, so that the scoring model to be used can perform translation scoring processing on all or part of the matching representation data, and obtain and output the model scoring result corresponding to the target translation text, so that the model scoring result can represent the score value obtained by the user's translated text relative to the target translation text.

[0173] Step 34: Determine the matching score corresponding to the target translation text according to the model scoring result corresponding to the target translation text.

[0174] The embodiment of the present application does not limit the implementation method of step 34. For example, the model scoring result corresponding to the target translation text can be directly determined as the matching score corresponding to the target translation text.

[0175] In fact, the semantics of text data obtained by arranging multiple words in different orders are different. Based on this, it can be seen that word order is very important for a text data. Therefore, in order to further improve the accuracy of translation scoring, the embodiment of the present application also provides a possible implementation of step 34. In this implementation, if the matching representation data corresponding to the target translation text includes the second representation data, then step 34 can specifically include steps 41-43:

[0176] Step 41: If it is determined that the second representation data corresponding to the target translation text is lower than the first representation threshold, a matching score corresponding to the target translation text is determined according to the model scoring result and the first preset scoring value.

[0177] The “first characterization threshold” may be preset, for example, the “first characterization threshold” may be 0.1.

[0178] The “first preset score value” is the highest score value corresponding to the preset left-closed and right-open interval [0, the first characterization threshold); and the embodiment of the present application does not limit the “first preset score value”, for example, it can be F base / 4.

[0179] Based on the relevant content of the above step 41, it can be known that after obtaining the second representation data corresponding to the target translation text, if it is determined that the second representation data is lower than the first representation threshold, it can be determined that the matching degree between the word order of the user translation text and the word order of the target translation text is very low, and thus it can be determined that the matching degree between the user translation text and the target translation text is also very low, so the matching score corresponding to the target translation text can be directly determined from [0, first preset score value] (for example, the matching score corresponding to the target translation text can be determined in the manner shown in the first line of formula (7) below). In this way, it can effectively avoid that the user translation text with an incorrect word order obtains a relatively high translation score, which is conducive to improving the accuracy of the translation score.

[0180] Step 42: If it is determined that the second representation data corresponding to the target translation text is not lower than the first representation threshold and lower than the second representation threshold, a matching score corresponding to the target translation text is determined according to the model scoring result and the second preset scoring value.

[0181] The “second characterization threshold” may be preset, for example, the “second characterization threshold” may be 0.2.

[0182] The “second preset score value” is the highest score value corresponding to the preset left-closed and right-open interval [first characterization threshold, second characterization threshold); and the embodiment of the present application does not limit the “second preset score value”, for example, it can be F base / 2.

[0183] Based on the relevant content of the above step 42, it can be known that after obtaining the second representation data corresponding to the target translation text, if it is determined that the second representation data is between the first representation threshold and the second representation threshold, it can be determined that the degree of match between the word order of the user translation text and the word order of the target translation text is relatively low, and thus it can be determined that the degree of match between the user translation text and the target translation text is also relatively low, so the matching score corresponding to the target translation text can be directly determined from [0, second preset score value] (for example, the matching score corresponding to the target translation text can be determined in the manner shown in the second line of formula (7) below). In this way, it can effectively avoid that the user translation text with many errors in word order obtains a relatively high translation score, which is conducive to improving the accuracy of the translation score.

[0184] Step 43: If it is determined that the second representation data corresponding to the target translation text is not lower than the second representation threshold, a matching score corresponding to the target translation text is determined according to the model scoring result.

[0185] In an embodiment of the present application, after obtaining the second characterization data corresponding to the target translation text, if it is determined that the second characterization data is not lower than the second characterization threshold, it can be determined that the degree of match between the word order of the user translation text and the word order of the target translation text is relatively high, and thus it can be determined that the degree of match between the user translation text and the target translation text is also relatively high, so the matching score corresponding to the target translation text can be determined with reference to the model scoring result (for example, the model scoring result is directly determined as the matching score corresponding to the target translation text. For another example, the product of the model scoring result and the "correction parameter value corresponding to the target translation text" below can be determined as the matching score corresponding to the target translation text).

[0186] Based on the relevant contents of the above steps 41 to 43, it can be known that the matching score corresponding to the target translation text can be determined by further referring to the matching degree between the word order of the user translation text and the word order of the target translation text, so that the matching score can more accurately represent the score value obtained by the user translation text relative to the target translation text.

[0187] In some cases, the relative amount of information carried by the user translation text and the amount of information carried by the target translation text can also characterize the degree of similarity between the user translation text and the target translation text. Therefore, in order to further improve the accuracy of the translation score, the present application embodiment also provides a possible implementation of step 34, which may specifically include steps 51-52:

[0188] Step 51: Determine a correction parameter value corresponding to the target translation text according to the similarity between the information quantity representation data of the user translation text and the information quantity representation data of the target translation text.

[0189] Among them, "information quantity characterizing data" is used to describe the information quantity of a text; and the embodiment of the present application does not limit the "information quantity characterizing data", for example, it may specifically include the text length and the number of words.

[0190] "Text length" is used to indicate the data length of a text data; and the embodiment of the present application does not limit the measurement method of "text length", for example, it can be measured by the number of characters in the text. For example, for the text "The air is very good today!", the text length of the text is 7.

[0191] "Number of words" is used to indicate the number of words included in a text data (for example, a user translation text or a target translation text); and the embodiments of the present application do not limit the method for determining the "number of words", for example, the text data may be first segmented to obtain a segmentation set of the text data; and then the segmentation set of the text data may be statistically processed to obtain the number of words in the text data.

[0192] In fact, since function words (such as prepositions, conjunctions, auxiliary words, modal particles) have basically no practical meaning, the semantic value of function words is relatively low. Therefore, in order to minimize the impact of function words on translation scores, the embodiment of the present application also provides a possible implementation method for determining the "number of words", which may specifically include: first performing word segmentation processing on the above-mentioned "text data" to obtain a word segmentation set of the text data; then filtering the word segmentation set of the text data to obtain a useful word set of the text data, so that the useful word set is used to record other words in the text data except useless words (for example, function words, etc.); finally, performing word number statistics processing on the useful word set to obtain the word number of the text data, so that the word number is used to represent the vocabulary with practical meaning in the text data.

[0193] The “correction parameter value corresponding to the target translation text” is used to characterize the similarity between the amount of information carried by the user translation text and the amount of information carried by the target translation text.

[0194] The embodiment of the present application does not limit step 51. For example, when the “information quantity representation data” includes text length and the number of words, step 51 may specifically include steps 511-3413:

[0195] Step 511: Determine a text length ratio according to the ratio between the text length of the user translation text and the text length of the target translation text.

[0196] The “text length ratio” is used to characterize the similarity in text length between the amount of information carried by the user translation text and the amount of information carried by the target translation text.

[0197] In addition, the embodiment of the present application does not limit the implementation method of step 5111. For example, it can be implemented using formula (4).

[0198]

[0199] In the formula, Indicates the above-mentioned "text length ratio"; l use Indicates the length of the text translated by the user; l target Indicates the text length of the target translation text.

[0200] Step 512: Determine a vocabulary ratio based on the ratio between the number of vocabulary in the user translation text and the number of vocabulary in the target translation text.

[0201] The “vocabulary ratio” is used to characterize the similarity in vocabulary between the amount of information carried by the user translation text and the amount of information carried by the target translation text.

[0202] In addition, the embodiment of the present application does not limit the implementation method of step 5112. For example, it can be implemented using formula (5).

[0203]

[0204] In the formula, represents the above-mentioned “ratio of vocabulary number”; W use Indicates the number of words in the user's translation text; W target Indicates the number of words in the target translation text.

[0205] Step 513: Determine the correction parameter value corresponding to the target translation text according to the sum of the text length ratio and the vocabulary number ratio.

[0206] In the embodiment of the present application, after obtaining the above-mentioned "text length ratio" and the above-mentioned "number of words ratio", the two ratios can be combined to determine the correction parameter value corresponding to the target translation text (for example, the sum of the text length ratio and the number of words ratio can be directly determined as the correction parameter value corresponding to the target translation text. For another example, the correction parameter value corresponding to the target translation text can be calculated using formula (6)), so that the correction parameter value can accurately characterize the difference between the amount of information carried by the user translation text and the amount of information carried by the target translation text.

[0207]

[0208] In the formula, Indicates the correction parameter value corresponding to the target translation text; Indicates the above-mentioned "text length ratio"; Indicates the above-mentioned “vocabulary number ratio”.

[0209] Based on the relevant content of the above step 51, it can be known that the correction parameter value corresponding to the target translated text can be determined according to the similarity between the information amount characterization data of the user translated text and the information amount characterization data of the target translated text, so that the correction parameter value can accurately characterize the similarity between the amount of information carried by the user translated text and the amount of information carried by the target translated text.

[0210] Step 52: Determine the matching score corresponding to the target translation text according to the product of the model score result corresponding to the target translation text and the correction parameter value corresponding to the target translation text.

[0211] The embodiment of the present application does not limit step 52. For example, it may specifically include: multiplying the product of the model score result corresponding to the target translation text and the correction parameter value corresponding to the target translation text to determine the matching score corresponding to the target translation text.

[0212] In fact, the semantics of text data obtained by arranging multiple words in different orders are different. Based on this, it can be seen that word order is very important for a text data. Therefore, in order to further improve the accuracy of translation scoring, the embodiment of the present application also provides a possible implementation of step 52, which can specifically include: according to the product between the model scoring result corresponding to the target translation text and the correction parameter value corresponding to the target translation text, and the above-mentioned "second representation data", determine the matching score corresponding to the target translation text (as shown in formula (7)).

[0213]

[0214] In the formula, Indicates the matching score corresponding to the target translation text; Indicates the model scoring result corresponding to the target translation text; represents the above-mentioned “second representation data” (for example, the second word segmentation matching result corresponding to the target translation text); Indicates the correction parameter value corresponding to the target translation text; F base Indicates a pre-set benchmark score (e.g., F base =2, or F base =10).

[0215] It should be noted that, for the relevant content of the above-mentioned “second characterization data”, please refer to the relevant content of the “second characterization data” in step 33 above.

[0216] Based on the relevant contents of the above steps 51 to 52, it can be known that the matching score corresponding to the target translation text can be determined by further referring to the relative size between the amount of information carried by the user translation text and the amount of information carried by the target translation text, so that the matching score can more accurately represent the score value obtained by the user translation text relative to the target translation text.

[0217] Based on the relevant contents of the above steps 31 to 34, it can be known that after obtaining the matching representation data corresponding to the target translation text, the matching representation data corresponding to the target translation text can be firstly processed by translation scoring model corresponding to the translation type of the text to be translated to obtain the model scoring result corresponding to the target translation text; and then the matching score corresponding to the target translation text is determined with reference to the model scoring result corresponding to the target translation text, so that the matching score can accurately represent the score value of the user translation text relative to the target translation text.

[0218] It should be noted that the matching score corresponding to any reference translation text can be implemented by any matching score determination process shown in the above steps 31 to 34. That is, the matching score determination process corresponding to the qth reference translation text can be implemented by any implementation method of the above “matching score determination process corresponding to the target translation text”, and it is only necessary to replace the “target translation text” with the “qth reference translation text” in any implementation method of the above “matching score determination process corresponding to the target translation text”. Wherein, q is a positive integer, q≤Q, Q is a positive integer, and Q represents the number of reference translation texts.

[0219] Based on the relevant content of S2 above, it can be known that after obtaining the matching representation data corresponding to each reference translation text, the matching scores corresponding to each reference translation text can be determined with reference to the matching representation data corresponding to the reference translation text, so that each matching score can accurately represent the score value of the user translation text relative to each reference translation text.

[0220] S3: Determine a translation score of the user translated text according to the matching score corresponding to at least one reference translated text.

[0221] The “translation score of the user translated text” is used to indicate the score value of the user translated text.

[0222] In addition, the embodiments of the present application do not limit the process of determining the "translation score of the user translated text". For example, if the number of reference translated texts is Q, the matching scores corresponding to the Q reference translated texts can be directly subjected to preset statistical processing (such as average calculation processing, maximum value screening processing, minimum value screening processing, etc.) to obtain the translation score of the user translated text.

[0223] In some cases, different scoring systems may be set for different test questions (for example, some test questions are 2-point system, some test questions are 10-point system, some test questions are 5-point system, etc.). Therefore, in order to improve the flexibility of translation scoring, the present application embodiment further provides a possible implementation of S3, which may specifically include S31-S32:

[0224] S31: Performing preset data processing on a matching score corresponding to at least one reference translation text to obtain a score to be used.

[0225] Among them, "preset data processing" refers to a pre-set data processing technology; and the embodiments of the present application are not limited to "preset data processing", for example, it can be an average calculation processing, a maximum value screening processing, or a minimum value screening processing, etc.

[0226] The "score to be used" is used to indicate the score value of the user's translated text under the above-mentioned "benchmark score".

[0227] S32: According to the preset full score of the text to be translated, the score to be used is converted into a score system to obtain the translation score of the user's translated text.

[0228] The "preset full score of the text to be translated" refers to the highest score (e.g., 2 points, 10 points, 5 points, etc.) that a user can get under the preset score system (e.g., 2 points, 10 points, 5 points, etc.) of the text to be translated.

[0229] The "score system conversion process" refers to mapping the score value under the benchmark score system (that is, the score system with the above-mentioned "base score" as the highest score) to the preset score system of the text to be translated.

[0230] In addition, the embodiment of the present application does not limit the implementation of S32. For example, it can be implemented using formula (8).

[0231]

[0232] In the formula, Grade use Indicates the translation score of the user's translated text; mark use Indicates the score to be used; Full base Indicates the above-mentioned "benchmark score"; Full question Indicates the preset maximum score for the text to be translated.

[0233] Based on the relevant contents of S1 to S3 above, it can be known that for the translation scoring method provided in the embodiment of the present application, after obtaining the user translation text input by the terminal user for the text to be translated, the user translation text is firstly matched with at least one reference translation text of the text to be translated respectively to obtain the matching representation data corresponding to each reference translation text; then, according to the matching representation data corresponding to each reference translation text, the matching score corresponding to each reference translation text is determined respectively; finally, according to the matching score corresponding to at least one reference translation text, the translation score of the user translation text is determined. Among them, because the above-mentioned "at least one reference translation text" can accurately represent the diversified translation results of the text to be translated, the matching score determined based on the "at least one reference translation text" can more accurately represent the matching degree between the user translation text and the diversified translation results of the text to be translated, so that the translation score determined based on these matching scores can more accurately represent the matching degree between the user translation text and the text to be translated, which is conducive to improving the accuracy of the translation score.

[0234] Based on the translation scoring method provided in the above method embodiment, the embodiment of the present application also provides a translation scoring device, which is explained and illustrated below in conjunction with the accompanying drawings.

[0235] Device Embodiment

[0236] The device embodiment introduces the translation scoring device. For related content, please refer to the above method embodiment.

[0237] See also Figure 3 , which is a structural schematic diagram of a translation scoring device provided in an embodiment of the present application.

[0238] The translation scoring device 300 provided in the embodiment of the present application includes:

[0239] The content matching unit 301 is used to, after obtaining a user translation text input by a terminal user for a text to be translated, perform preset matching processing on the user translation text and at least one reference translation text of the text to be translated, to obtain matching representation data corresponding to each reference translation text;

[0240] A translation scoring unit 302, configured to determine a matching score corresponding to each of the reference translation texts according to the matching representation data corresponding to each of the reference translation texts;

[0241] The score determination unit 303 is configured to determine the translation score of the user translated text according to the matching score corresponding to the at least one reference translated text.

[0242] In a possible implementation manner, the at least one reference translation text includes a target translation text, and the content matching unit 301 includes:

[0243] A word matching subunit, used for performing word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text;

[0244] A word matching subunit, used for performing word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text;

[0245] The result collection subunit is used to merge the character matching result with the word matching result to obtain matching representation data corresponding to the target translation text.

[0246] In a possible implementation, the word matching subunit is specifically used to: pre-process the user translation text to obtain a word segmentation set of the user translation text; pre-process the target translation text to obtain a word segmentation set of the target translation text; perform a first word matching process on the word segmentation set of the user translation text and the word segmentation set of the target translation text to obtain a first word segmentation matching result; perform a second word matching process on the word segmentation set of the user translation text and the word segmentation set of the target translation text to obtain a second word segmentation matching result; merge the first word segmentation matching result and the second word segmentation matching result to obtain a word matching result corresponding to the target translation text.

[0247] In a possible implementation, the preprocessing includes word segmentation processing and filtering processing.

[0248] In a possible implementation, the at least one reference translation text includes a target translation text, and the translation scoring unit 302 includes:

[0249] A type determination subunit, used to determine the translation type of the text to be translated;

[0250] A model search subunit, configured to search the to-be-used scoring model corresponding to the translation type from the preset mapping relationship; wherein the preset mapping relationship includes a correspondence relationship between the translation type and the pre-constructed to-be-used scoring model;

[0251] A model scoring subunit is used to score the matching representation data corresponding to the target translation text using the scoring model to be used, so as to obtain a model scoring result corresponding to the target translation text;

[0252] The score determination subunit is used to determine the matching score corresponding to the target translation text according to the model scoring result.

[0253] In a possible implementation, the matching representation data corresponding to the target translation text includes first representation data and second representation data; wherein the first representation data is used to represent a word matching result between the user translation text and the target translation text; and the second representation data is used to represent a word order matching result between the user translation text and the target translation text;

[0254] The model scoring subunit is specifically used for:

[0255] If it is determined that the translation type of the text to be translated is a sentence translation type, the first representation data is input into the scoring model to be used to obtain a model scoring result corresponding to the target translated text output by the scoring model to be used; if it is determined that the translation type of the text to be translated is a short text translation type, the matching representation data corresponding to the target translated text is input into the scoring model to be used to obtain a model scoring result corresponding to the target translated text output by the scoring model to be used.

[0256] In a possible implementation manner, the score determination subunit includes:

[0257] A parameter determination subunit, configured to determine a correction parameter value corresponding to the target translation text according to a similarity between the information quantity characterization data of the user translation text and the information quantity characterization data of the target translation text;

[0258] The score correction subunit is used to determine the matching score corresponding to the target translation text according to the product between the model score result and the correction parameter value.

[0259] In a possible implementation manner, the information quantity characterization data includes text length and the number of words, and the parameter determination subunit is specifically used to:

[0260] A text length ratio is determined according to a ratio between a text length of the user translated text and a text length of the target translated text; a vocabulary number ratio is determined according to a ratio between a number of words in the user translated text and a number of words in the target translated text; and a correction parameter value corresponding to the target translated text is determined according to a sum of the text length ratio and the vocabulary number ratio.

[0261] In a possible implementation manner, the matching representation data corresponding to the target translation text includes second representation data; wherein the second representation data is used to represent a word order matching result between the user translation text and the target translation text;

[0262] The score determination subunit is specifically used to: if it is determined that the second representation data is lower than the first representation threshold, determine the matching score corresponding to the target translation text according to the model scoring result and the first preset scoring value; if it is determined that the second representation data is not lower than the first representation threshold and is lower than the second representation threshold, determine the matching score corresponding to the target translation text according to the model scoring result and the second preset scoring value; if it is determined that the second representation data is not lower than the second representation threshold, determine the matching score corresponding to the target translation text according to the model scoring result.

[0263] In a possible implementation, the score determination unit 303 is specifically used to: perform preset data processing on the matching score corresponding to the at least one reference translation text to obtain a score to be used; and perform point conversion processing on the score to be used according to a preset full score value of the text to be translated to obtain a translation score of the user translated text.

[0264] In a possible implementation manner, the translation scoring device 300 further includes:

[0265] The reference generation unit is used to obtain a standard translation text of the text to be translated; perform word segmentation processing on the standard translation text to obtain a vocabulary set of the standard translation text; search for a synonym of at least one word in the vocabulary set from a pre-built synonym database; and generate at least one reference translation text of the text to be translated based on the standard translation text and the synonym of the at least one word.

[0266] Furthermore, the embodiment of the present application also provides a translation scoring device, including: a processor, a memory, and a system bus;

[0267] The processor and the memory are connected via the system bus;

[0268] The memory is used to store one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes any one of the implementation methods of the above-mentioned translation scoring method.

[0269] Furthermore, an embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any one of the implementation methods of the above-mentioned translation scoring method.

[0270] Furthermore, an embodiment of the present application also provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any one of the implementation methods of the above-mentioned translation scoring method.

[0271] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., including several instructions for enabling a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0272] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0273] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0274] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A translation scoring method, It is characterized in that The method comprises: After obtaining a user translation text input by a terminal user for the text to be translated, performing a preset matching process on the user translation text and at least one reference translation text of the text to be translated, respectively, to obtain matching representation data corresponding to each reference translation text; Determining the matching scores of the reference translation texts according to the matching representation data corresponding to the reference translation texts; Determining a translation score of the user translated text according to a matching score corresponding to the at least one reference translated text; The at least one reference translation text includes a target translation text, and the process of determining matching representation data corresponding to the target translation text includes: Performing word matching processing on the user translated text and the target translated text to obtain a word matching result corresponding to the target translated text; Performing word matching processing on the user translated text and the target translated text to obtain a word matching result corresponding to the target translated text; The character matching result and the word matching result are combined to obtain matching representation data corresponding to the target translation text.

2. The method according to claim 1, It is characterized in that The process of determining the word matching result includes: Preprocessing the user translated text to obtain a word segmentation set of the user translated text; Preprocessing the target translation text to obtain a word set of the target translation text; Performing a first word matching process on the word set of the user translation text and the word set of the target translation text to obtain a first word matching result; Performing a second word matching process on the word set of the user translation text and the word set of the target translation text to obtain a second word matching result; The first word segmentation matching result and the second word segmentation matching result are combined to obtain a word matching result corresponding to the target translation text.

3. The method according to claim 2, It is characterized in that The preprocessing includes word segmentation processing and filtering processing.

4. The method according to claim 1, It is characterized in that The at least one reference translation text includes a target translation text, and a process of determining a matching score corresponding to the target translation text includes: Determining the translation type of the text to be translated; Searching for a scoring model to be used corresponding to the translation type from a preset mapping relationship; wherein the preset mapping relationship includes a corresponding relationship between the translation type and the pre-constructed scoring model to be used; Using the scoring model to be used, scoring the matching representation data corresponding to the target translation text to obtain a model scoring result corresponding to the target translation text; According to the model scoring result, a matching score corresponding to the target translation text is determined.

5. The method according to claim 4, It is characterized in that The matching representation data corresponding to the target translation text includes first representation data and second representation data; wherein the first representation data is used to represent the word matching result between the user translation text and the target translation text; and the second representation data is used to represent the word order matching result between the user translation text and the target translation text; The process of determining the model scoring result includes: If it is determined that the translation type of the text to be translated is a sentence translation type, inputting the first representation data into the scoring model to be used to obtain a model scoring result corresponding to the target translation text output by the scoring model to be used; If it is determined that the translation type of the text to be translated is a short text translation type, the matching representation data corresponding to the target translation text is input into the scoring model to be used to obtain a model scoring result corresponding to the target translation text output by the scoring model to be used.

6. The method according to claim 4, It is characterized in that The process of determining the matching score includes: Determining a correction parameter value corresponding to the target translated text according to a similarity between the information quantity characterization data of the user translated text and the information quantity characterization data of the target translated text; The matching score corresponding to the target translation text is determined according to the product of the model scoring result and the correction parameter value.

7. The method according to claim 6, It is characterized in that The information quantity characterization data includes text length and word count, and the process of determining the correction parameter value includes: Determining a text length ratio according to a ratio between a text length of the user translated text and a text length of the target translated text; Determining a vocabulary ratio according to a ratio between the number of vocabulary in the user translation text and the number of vocabulary in the target translation text; The correction parameter value corresponding to the target translation text is determined according to the sum of the text length ratio and the vocabulary number ratio.

8. The method according to claim 4, It is characterized in that The matching representation data corresponding to the target translation text includes second representation data; wherein the second representation data is used to represent the word order matching result between the user translation text and the target translation text; The process of determining the matching score includes: If it is determined that the second representation data is lower than the first representation threshold, determining a matching score corresponding to the target translation text according to the model scoring result and a first preset scoring value; If it is determined that the second representation data is not lower than the first representation threshold and lower than the second representation threshold, determining a matching score corresponding to the target translation text according to the model scoring result and a second preset scoring value; If it is determined that the second representation data is not lower than a second representation threshold, a matching score corresponding to the target translation text is determined according to the model scoring result.

9. The method according to claim 1, It is characterized in that The process of determining the translation score includes: Performing preset data processing on the matching score corresponding to the at least one reference translation text to obtain a score to be used; According to the preset full score of the text to be translated, the score to be used is converted into a score system to obtain the translation score of the user translated text.

10. The method according to claim 1, It is characterized in that The process of generating the at least one reference translation text comprises: Obtaining a standard translation text of the text to be translated; Performing word segmentation processing on the standard translation text to obtain a vocabulary set of the standard translation text; Searching for a synonym for at least one word in the vocabulary from a pre-built synonym database; At least one reference translation text of the text to be translated is generated according to the standard translation text and the synonym of the at least one vocabulary.

11. A translation scoring device, It is characterized in that include: A content matching unit, configured to, after acquiring a user translation text input by a terminal user for a text to be translated, perform preset matching processing on the user translation text and at least one reference translation text of the text to be translated, respectively, to obtain matching representation data corresponding to each reference translation text; A translation scoring unit, configured to determine a matching score corresponding to each of the reference translation texts according to the matching representation data corresponding to each of the reference translation texts; a score determination unit, configured to determine a translation score of the user translated text according to a matching score corresponding to the at least one reference translated text; The at least one reference translation text includes a target translation text, and the content matching unit includes: A word matching subunit, used for performing word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text; A word matching subunit, used for performing word matching processing on the user translation text and the target translation text to obtain a word matching result corresponding to the target translation text; The result collection subunit is used to merge the character matching result with the word matching result to obtain matching representation data corresponding to the target translation text.

12. A translation scoring device, It is characterized in that The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 10.

13. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the method according to any one of claims 1 to 10.

14. A computer program product, It is characterized in that When the computer program product is executed on a terminal device, the terminal device executes the method according to any one of claims 1 to 10.

Images