Translation Quality Assessment Method, Device, Electronic Device and Storage Medium

Abstract

The present invention provides a method, apparatus, electronic device and storage medium for translation quality assessment. The method includes: determining a source language speech and a target language text; performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech assessment result based on the source language standard text; determining a text assessment result based on the target language text; and determining the translation quality of the target language text based on the speech assessment result and the text assessment result. The translation quality assessment method, apparatus, electronic device and storage medium provided by the embodiments of the present invention take into account both the quality of the source language speech and the quality of the target language text, thereby improving the accuracy of translation quality assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language, and particularly to a method, apparatus, electronic device and storage medium for evaluating translation quality. Background Art

[0002] With the further expansion of the market demand for translation, there is a strong demand for instant translation, and people need instant translation for direct communication. However, the current machine translation ability still fails to reach the translation level of professional translators. Therefore, how to accurately evaluate and measure the translation quality of instant translation in different scenarios has become increasingly important.

[0003] The existing evaluation of instant translation quality mainly focuses on evaluating the translated text, and determines the translation quality according to the similarity between the text of the original speech and the translated text. However, due to the diversity of language expressions and the richness of language directions, the credibility of translation quality evaluation is still insufficient. Summary of the Invention

[0004] The present invention provides a method, apparatus, electronic device and storage medium for evaluating translation quality, so as to solve the defect of poor credibility of translation quality evaluation in the prior art.

[0005] The present invention provides a method for evaluating translation quality, including:

[0006] Determine the source language speech and the target language text;

[0007] Perform source language standard translation on the source language speech to obtain a source language standard text, and determine a speech evaluation result based on the source language standard text;

[0008] Determine a text evaluation result based on the target language text;

[0009] Determine the translation quality of the target language text based on the speech evaluation result and the text evaluation result.

[0010] According to the method for evaluating translation quality provided by the present invention, the determining of the speech evaluation result based on the source language standard text includes:

[0011] Determine a word-level speech evaluation result based on the translation accuracy rate of the word segmentation in the source language standard text for the source language speech;

[0012] And / or, determine a sentence-level speech evaluation result based on the semantic accuracy and / or expression fluency of the sentence clauses in the source language standard text;

[0013] Determine the speech evaluation result based on the word-level speech evaluation result and / or the sentence-level speech evaluation result.

[0014] According to the translation quality evaluation method provided by the present invention, the translation accuracy rate of the word segmentation in the source language standard text for the source language speech is determined based on the following steps:

[0015] Based on the technical field of the word segmentation in the transcription text of the source language speech, determine the first technical field;

[0016] Based on the first technical field, determine the translation accuracy rate of the word segmentation in the source language standard text for the source language speech.

[0017] According to the translation quality evaluation method provided by the present invention, determining the text evaluation result based on the target language text includes:

[0018] Based on the translation accuracy rate of the word segmentation in the target language text for the source language standard text, determine the word-level text evaluation result;

[0019] And / or, perform source language standardization translation on the target language text to obtain a translation standard text, and based on the semantic accuracy and / or expression fluency of the clauses in the translation standard text, determine the sentence-level text evaluation result;

[0020] Based on the word-level text evaluation result and / or the sentence-level text evaluation result, determine the text evaluation result.

[0021] According to the translation quality evaluation method provided by the present invention, the translation accuracy rate of the word segmentation in the target language text for the source language standard text is determined based on the following steps:

[0022] Based on the technical field of the word segmentation in the target language text, determine the second technical field;

[0023] Based on the second technical field, determine the translation accuracy rate of the word segmentation in the target language text for the source language standard text.

[0024] According to the translation quality evaluation method provided by the present invention, the semantic accuracy of the clause is determined based on the following steps:

[0025] Determine the target standard word order, and the target standard word order is determined based on the standard word order of the source language in the first technical field and / or the standard word order of the target language in the second technical field;

[0026] Based on the target standard word order, adjust the word order of the clauses in the source language standard text or the translation standard text, and obtain the semantic accuracy of the clauses in the source language standard text or the translation standard text after the word order adjustment.

[0027] According to the translation quality evaluation method provided by the present invention, the expression fluency of the clause is determined based on the following steps:

[0028] Based on the relevance of word segmentation in clauses of the source language standard text or the translation standard text, determine the expression fluency of the clauses in the source language standard text or the translation standard text.

[0029] The present invention also provides a translation quality evaluation device, including:

[0030] A voice and text determination unit, configured to determine the source language voice and the target language text;

[0031] A voice evaluation unit, configured to perform source language standard translation on the source language voice to obtain a source language standard text, and determine a voice evaluation result based on the source language standard text;

[0032] A text evaluation unit, configured to determine a text evaluation result based on the target language text;

[0033] A translation quality determination unit, configured to determine the translation quality of the target language text based on the voice evaluation result and the text evaluation result.

[0034] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the translation quality evaluation method described in any one of the above is implemented.

[0035] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the translation quality evaluation method described in any one of the above is implemented.

[0036] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the steps of the translation quality evaluation method described in any one of the above are implemented.

[0037] The translation quality evaluation method, device, electronic device, and storage medium provided by the present invention perform source language standard translation on the source language voice to obtain a source language standard text, determine a voice evaluation result based on the source language standard text; and determine a text evaluation result based on the target language text; and then determine the translation quality of the target language text based on the voice evaluation result and the text evaluation result. Since the quality of the source language voice and the quality of the target language text are both considered, the accuracy of translation quality evaluation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0039] Figure 1 It is a schematic flowchart of the translation quality evaluation method provided by the present invention;

[0040] Figure 2 It is a schematic flowchart of the speech evaluation method provided by the present invention;

[0041] Figure 3 It is one of the schematic flowcharts of the method for determining translation accuracy provided by the present invention;

[0042] Figure 4 It is a schematic flowchart of step 130 in the translation quality evaluation method provided by the present invention;

[0043] Figure 5 It is another schematic flowchart of the method for determining translation accuracy provided by the present invention;

[0044] Figure 6 It is a schematic flowchart of the method for determining the semantic accuracy of clause segmentation provided by the present invention;

[0045] Figure 7 It is a schematic structural diagram of the translation quality evaluation device provided by the present invention;

[0046] Figure 8 It is a schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0047] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0048] The existing evaluation of instant translation quality mainly focuses on evaluating the translated text, and determines the translation quality according to the similarity between the text of the original speech and the translated text. However, due to the diversity of language expressions and the richness of language directions, the credibility of translation quality evaluation is still insufficient.

[0049] In response to this, the embodiments of the present invention provide a translation quality evaluation method. Figure 1It is a schematic flowchart of the translation quality evaluation method provided by the present invention. As Figure 1 shown, the method includes:

[0050] Step 110, determining the source language speech and the target language text.

[0051] Specifically, the source language speech is the speech data used for translation. The source language speech can be obtained through a sound pickup device. Here, the sound pickup device can be a smart phone, a tablet computer, or also a smart appliance such as a speaker, a TV, and an air conditioner, etc. After the sound pickup device picks up the source language speech through the microphone array, it can also amplify and reduce the noise of the source language speech. In addition, the source language speech can be a speech segment formed after the sound pickup ends, or also a speech stream during the real-time sound pickup process. The embodiments of the present invention do not make specific limitations thereto.

[0052] The target language text is the text to be evaluated of the translation that needs to be evaluated for quality. The target language text is obtained by translating the source language speech. The target language text can be a translation obtained by machine translation or also a translation obtained by a translator's manual translation. The embodiments of the present invention do not make specific limitations thereto.

[0053] The source language and the target language here can be languages in different countries and regions. For example, speech translation can support multiple languages, such as Mandarin, Japanese, Russian, English, etc. Among them, the source language speech can also support various dialects, such as various dialects corresponding to the Mandarin language, such as Sichuan dialect, Cantonese, Minnan dialect, Uyghur, Shanghai dialect, etc.

[0054] Step 120, performing source language standardized translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text.

[0055] Specifically, considering the diversity of language expressions, taking Mandarin as the source language as an example, since the standard degree of Mandarin of people in each region is different, the input speech will be more or less with a dialect accent, which will affect the translation quality of the target language text; and everyone's language expression habits are also different, which will also affect the translation quality of the target language text. Therefore, the translation quality evaluation method provided by the embodiments of the present invention takes into account the speech quality of the input speech, and can also perform speech evaluation on the source language speech before comprehensively evaluating the translation quality.

[0056] Performing source language standardized translation on the source language speech means translating the source language speech into the corresponding standard language to obtain a source language standard text. For example, Sichuan dialect speech or non-standard Mandarin speech can be translated into standard Mandarin text. At this time, the obtained standard Mandarin text is the source language standard text.

[0057] Subsequently, the source language speech is evaluated according to the obtained source language standard text to obtain a speech evaluation result. Since the source language standard text only involves the standardization correction of the source language speech in terms of accent, the speech evaluation result based on the source language standard text can directly reflect the speech quality of the source language speech after removing accent interference. For example, it can reflect whether the meaning of the source language speech itself is clear and whether the pronunciation is ambiguous.

[0058] The speech evaluation result here can start from the semantics of the source language speech and the source language standard text after word segmentation, and evaluate the accuracy of the source language speech in terms of words, taking word segmentation as the unit, so as to obtain the word-level evaluation result of the speech; it can also start from the overall semantics of the source language speech and the source language standard text after sentence segmentation, and evaluate the semantic accuracy and / or expression fluency of the source language speech in terms of sentences, taking sentence segmentation as the unit, so as to obtain the sentence-level evaluation result of the speech.

[0059] Step 130: Determine a text evaluation result based on the target language text.

[0060] Specifically, the target language text can be evaluated to obtain a text evaluation result, and the text evaluation result can characterize the translation quality of the target language text. The text evaluation result here can start from the semantics of the source language standard text and the target language text after word segmentation, and evaluate the translation accuracy of the words in the target language text in terms of words, taking word segmentation as the unit, so as to obtain the word-level evaluation result of the text; it can also start from the overall semantics of the source language standard text and the target language text after sentence segmentation, and evaluate the semantic accuracy and / or expression fluency of the sentences in the target language text in terms of sentences, taking sentence segmentation as the unit, so as to obtain the sentence-level evaluation result of the text.

[0061] Step 140: Determine the translation quality of the target language text based on the speech evaluation result and the text evaluation result.

[0062] Specifically, the speech evaluation result and the text evaluation result can be fused to obtain a translation quality evaluation result of the target language text. Here, corresponding weights can be set for the speech evaluation result and the text evaluation result, and then the speech evaluation result and the text evaluation result are weighted and fused. It is also possible to fuse the speech evaluation result and the text evaluation result through a multi-layer perceptron network or other forms of fusion networks. The embodiments of the present invention do not make specific limitations on this.

[0063] Furthermore, the word-level evaluation result and / or the sentence-level evaluation result of the speech obtained by evaluating the quality from different granularities, and the word-level evaluation result and / or the sentence-level evaluation result of the text obtained by evaluating the quality from different granularities can be fused to further improve the accuracy of the translation quality evaluation.

[0064] The method provided by the embodiments of the present invention performs source language standard translation on the source language speech to obtain a source language standard text, determines a speech evaluation result based on the source language standard text; and determines a text evaluation result based on the target language text; then determines the translation quality of the target language text based on the speech evaluation result and the text evaluation result. Since both the quality of the source language speech and the quality of the target language text are considered, the accuracy of translation quality evaluation is improved.

[0065] Based on the above embodiments, Figure 2 is a schematic flowchart of the speech evaluation method provided by the present invention. As Figure 2 shown, determining the speech evaluation result based on the source language standard text in step 120 includes:

[0066] Step 121, determining a word-level speech evaluation result based on the translation accuracy rate of the word segmentation in the source language standard text for the source language speech;

[0067] And / or, step 122, determining a sentence-level speech evaluation result based on the semantic accuracy and / or expression fluency of the sentence segmentation in the source language standard text;

[0068] Step 123, determining the speech evaluation result based on the word-level speech evaluation result and / or the sentence-level speech evaluation result.

[0069] Specifically, considering that the pronunciation of the input speech person may not be very standard, for example, with a local accent, or the input speech contains domain vocabulary in a certain field, which may affect the translation quality evaluation result of the target language text. Therefore, the quality of the source language speech can be evaluated from different granularities. The speech evaluation result can start from the semantics of the source language speech and the word segmentation of the source language standard text, and evaluate the translation accuracy rate of the word segmentation in the source language standard text for the source language speech in units of word segmentation, so as to obtain the word-level speech evaluation result.

[0070] Among them, the translation accuracy rate can represent the possibility of translating the speech word segmentation into the standard text word segmentation at the corresponding position when sequentially translating the word segmentation in the source language speech.

[0071] It should be noted that here, the word-level speech evaluation result can be determined based on the translation accuracy rate of each word segmentation in the source language speech and the source language standard text; or the word-level speech evaluation result can be determined based on some word segmentation in the source language speech and the source language standard text, such as some key word segmentation or domain vocabulary. The embodiments of the present invention do not make specific limitations on this.

[0072] For example, if the source language speech is "What are you doing? Now", corresponding to the Chinese expression habits and grammatical structures, the corresponding standard text expression in the source language is "What are you doing now?". Then, based on the word segmentation in the source language standard text, namely "you", "now", "are", "doing what", the translation accuracy of each word segmentation in the source language speech, namely "you", "are", "doing what", "now", can be determined, thereby determining the word-level speech evaluation result. The higher the translation accuracy of the word segmentation, the better the word-level speech evaluation result; the lower the translation accuracy of the word segmentation, the worse the word-level speech evaluation result.

[0073] The speech evaluation result can also start from the overall semantics of the clauses in the source language standard text, taking the clause as the unit, and evaluate the semantic accuracy and / or expression fluency of the source language speech, thereby obtaining the sentence-level evaluation result of the speech.

[0074] Among them, the semantic accuracy can characterize the clarity of the semantics expressed by the source language speech. The higher the semantic accuracy, the clearer the semantics of the source language speech, the smaller the possibility of ambiguity, and the better the sentence-level evaluation result of the speech; the lower the semantic accuracy, the less clear the semantics of the source language speech, the greater the possibility of ambiguity, and the worse the sentence-level evaluation result of the speech. For example, the source language speech is: "Where are you now?" or "Where are you? Now", both of which can clearly express the accurate meaning of the speech; while "you now" belongs to the source language speech with lower semantic accuracy and cannot clearly express the accurate meaning of the speech, which will affect the translation evaluation result of the target language.

[0075] The expression fluency can characterize the fluency of the overall expression of the source language speech, reflecting the degree to which the expression of the source language speech conforms to the expression habits of the corresponding source language. For example, if the expression in the source language speech is smooth, with few pauses, and the word order of the subject, verb, and object conforms to the expression habits of the corresponding language, etc., it can be considered that the expression fluency is relatively high. The higher the expression fluency, the better the sentence-level evaluation result of the speech; the lower the expression fluency, the worse the sentence-level evaluation result of the speech. It should be noted that the sentence-level speech evaluation result can be determined based on the semantic accuracy and / or expression fluency of each clause in the source language standard text; or it can be determined based on some clauses in the source language standard text, such as the semantic accuracy and / or expression fluency of some key clauses.

[0076] Then, determine the speech evaluation result according to the word-level speech evaluation result, or determine the speech evaluation result according to the sentence-level speech evaluation result; or fuse the word-level speech evaluation result and the sentence-level speech evaluation result to determine the speech evaluation result.

[0077] The method provided by the embodiment of the present invention determines the speech evaluation result based on the word-level speech evaluation result and / or the sentence-level speech evaluation result, evaluates the speech quality from different dimensions, and thus improves the credibility of the translation quality evaluation of the target language text.

[0078] Based on any one of the above embodiments, Figure 3 is one of the schematic flowcharts of the translation accuracy determination method provided by the present invention. As Figure 3 shown, the translation accuracy of the word segmentation in the source language standard text in step 121 for the source language speech is determined based on the following steps:

[0079] Step 310, determine the first technical field based on the technical field of the word segmentation in the transcription text of the source language speech;

[0080] Step 320, determine the translation accuracy of the word segmentation in the source language standard text for the source language speech based on the first technical field.

[0081] Specifically, considering that each technical field has domain vocabulary, under different fields, the same speech word segmentation may correspond to different translation texts. Therefore, when evaluating the translation quality of the target language text, it can be evaluated based on the technical field to which the source language speech belongs.

[0082] The transcription text of the source language speech is the text obtained by performing speech transcription on the source language speech. Among them, a speech recognition model can be used for speech transcription. The source language speech is input into the speech recognition model to obtain the transcription text of the source language speech.

[0083] The technical field of the word segmentation in the transcription text can be determined by taking the word segmentation as a unit according to the probability of the word segmentation appearing in the domain vocabulary. For example, the domain vocabulary can include daily conversations, business communications, and academic exchanges. The domain vocabulary with the highest probability of the word segmentation appearing in it is used as the technical field of the word segmentation. The first technical field is the technical field determined based on the technical field of the word segmentation in the transcription text of the source language speech.

[0084] For example, the transcription text of the source language speech is: Where are you? Now. The corresponding word segmentations are: you, in, where, now, and thus the first technical field can be determined as the "daily conversation" field.

[0085] Under the technical field corresponding to the word segmentation, determine the translation accuracy of the word segmentation in the source language standard text for the source language speech. Specifically, under the technical field corresponding to the word segmentation, perform standardized translation decoding on the source language standard text to obtain the translation accuracy of the word segmentation in the transcription text of the source language speech. The word-level speech evaluation result can be the sum of the translation accuracies of the word segmentations in the transcription text of the source language speech.

[0086] The method provided by the embodiment of the present invention further improves the credibility of the word-level speech evaluation result by determining the translation accuracy of the word segmentation in the source language standard text to the source language speech based on the first technical field.

[0087] Based on any of the above embodiments, Figure 4 is a flow chart of step 130 in the translation quality assessment method provided by the present invention, such as Figure 4 As shown, step 130 specifically includes:

[0088] Step 131, determining a word-level text evaluation result based on the translation accuracy of the segmented words in the target language text to the source language standard text;

[0089] And / or, step 132, performing a standardized translation of the target language text into the source language to obtain a translation standard text, and determining a sentence-level text evaluation result based on the semantic accuracy and / or expression fluency of the sentences in the translation standard text;

[0090] Step 133: Determine a text evaluation result based on the word-level text evaluation result and / or the sentence-level text evaluation result.

[0091] Specifically, for the target language text, the quality of the target language text can also be evaluated at different granularities.

[0092] The text evaluation result can be based on the semantics of the target language text and the source language standard text segmentation, and evaluate the translation accuracy of the target language text segmentation to the source language standard text with the segmentation as the unit, so as to obtain the word-level text evaluation result. Among them, the translation accuracy can represent the possibility of translating the standard text segmentation into the target language text segmentation at the corresponding position when translating the segmentation in the source language standard text in sequence. For example, for the Chinese "你是谁", the corresponding English expression is "who are you", and the word-level text evaluation result can be determined based on the translation accuracy of "你" to "you", the translation accuracy of "是" to "are", and the translation accuracy of "谁" to "who".

[0093] The target language text can also be standardized into the source language to obtain a translation standard text. Starting from the overall semantics of the sentences in the translation standard text, the semantic accuracy and / or expression fluency of the sentences in the translation standard text can be evaluated on a sentence-by-sentence basis to obtain a sentence-level evaluation result of the text.

[0094] The text evaluation result may be determined based on the word-level text evaluation result, or based on the sentence-level text evaluation result; or the word-level text evaluation result and the sentence-level text evaluation result may be integrated to determine the text evaluation result.

[0095] The method provided by the embodiments of the present invention determines the text evaluation result based on the word-level text evaluation result and / or the sentence-level text evaluation result, evaluates the text quality from different dimensions, and thus improves the credibility of the translation quality evaluation of the target language text.

[0096] Based on any of the above embodiments, Figure 5 is the second schematic flow chart of the method for determining the translation accuracy provided by the present invention. As Figure 5 shown, in step 131, the translation accuracy of the word segmentation in the target language text for the source language standard text is determined based on the following steps:

[0097] Step 510, determine the second technical field based on the technical field of the word segmentation in the target language text;

[0098] Step 520, determine the translation accuracy of the word segmentation in the target language text for the source language standard text based on the second technical field.

[0099] Specifically, in the process of determining the translation accuracy of the word segmentation in the target language text for the source language standard text, the technical field of the word segmentation in the target language text is also considered, and the second technical field is determined according to the technical field of the word segmentation in the target language text.

[0100] It should be noted that since the judgment bases of the first technical field and the second technical field are different, the obtained second technical field may be the same as or different from the first technical field.

[0101] Then, under the second technical field, based on the coding features of the word segmentation in the target language text, the target language text is translation-decoded to obtain the translation accuracy of the word segmentation in the target language text for the source language standard text. The word-level text evaluation result can be the sum of the translation accuracies of the word segmentation in the target language text for the source language standard text.

[0102] The method provided by the embodiments of the present invention determines the translation accuracy of the word segmentation in the target language text for the source language standard text based on the second technical field, and further improves the credibility of the evaluation result of the target language text.

[0103] Based on any of the above embodiments, Figure 6 is the schematic flow chart of the method for determining the semantic accuracy of sentence segmentation provided by the present invention. As Figure 6 shown, the semantic accuracy of sentence segmentation is determined based on the following steps:

[0104] Step 610, determine the target standard word order, where the target standard word order is determined based on the standard word order of the source language under the first technical field and / or the standard word order of the target language under the second technical field;

[0105] Step 620: Based on the target standard word order, adjust the word order of the clauses in the source language standard text or the translated standard text, and obtain the semantic accuracy of the clauses in the source language standard text or the translated standard text after the word order adjustment.

[0106] Specifically, the first technical field is the technical field determined based on the technical field of word segmentation in the transcription text of the source language speech. Under the technical field to which it belongs, determine the standard word order of the source language. Among them, the standard word order refers to the word order in which the word segmentation in the transcription text of the source language speech conforms to the expression habits of the source language. For example, determine the standard word order according to the subject, predicate, object, adverbial, etc. of the Chinese language.

[0107] The second technical field is the technical field determined based on the technical field of word segmentation in the target language text. Under the technical field to which it belongs, determine the standard word order of the target language. For example, if the target language is English, determine the standard word order of the target language according to the grammar and expression habits of English.

[0108] Based on the target standard word order, adjust the word order of each clause in the source language standard text obtained by translating the source language speech, or adjust the word order of each clause in the translated standard text obtained by translating the target language text, and determine the semantic accuracy of the clauses in the source language standard text or the translated standard text according to the source language standard text or the translated standard text after the word order adjustment.

[0109] Based on any of the above embodiments, the expression fluency of the clause is determined based on the following steps: Determine the expression fluency of the clause in the source language standard text or the translated standard text based on the relevance of the word segmentation in the clause of the source language standard text or the translated standard text.

[0110] Specifically, determine the relevance of the word segmentation in the clause of the source language standard text or the translated standard text, so as to determine the expression fluency of the clause in the source language standard text or the translated standard text based on the relevance between the word segments. Among them, the higher the relevance of the word segmentation in the clause, the better the semantic coherence of the word segmentation in the clause, and the higher the expression fluency of the clause.

[0111] Furthermore, the determination of the expression fluency can be achieved through a neural network model. First, determine the word vectors of the word segmentation in the clause of the source language standard text or the translated standard text, and use the attention mechanism to determine the relevance of the word segmentation in the clause of the source language standard text or the translated standard text, so as to encode and obtain the semantic encoding vector of the word segmentation in the clause of the source language standard text or the translated standard text. Subsequently, use a scoring network, such as a softmax layer, to score the text semantic vector obtained by fusing the semantic encoding vectors of the word segmentation, and obtain the expression fluency of the clause.

[0112] Based on any of the above embodiments, taking the source language as Mandarin as an example, the present invention provides a translation quality evaluation method, and the method includes:

[0113] (1) Determine the original speech and the text in the target language;

[0114] (2) Perform Mandarin standardization translation on the original speech to obtain the Mandarin standard text; perform Mandarin standardization translation on the text in the target language to obtain the translation standard text;

[0115] (3) Based on the technical field of each word segment in the transcription text of the original speech, determine the first technical field; under the first technical field, determine the translation accuracy rate of each word segment in the Mandarin standard text for the original speech as the word-level speech evaluation result, denoted as A 语 ;

[0116] (4) Based on the technical field of each word segment in the text in the target language, determine the second technical field; under the second technical field, determine the translation accuracy rate of each word segment in the text in the target language for the Mandarin standard text as the word-level text evaluation result, denoted as A 译 ;

[0117] (5) Under the first technical field, determine the semantic accuracy of each clause in the Mandarin standard text, denoted as B1; determine the expression fluency of each clause in the Mandarin standard text, denoted as B2; then the sentence-level speech evaluation result can be denoted as: B 语 = (B1 + B2) / 2;

[0118] (6) Under the second technical field, determine the semantic accuracy of each clause in the translation standard text, denoted as B3; determine the expression fluency of each clause in the translation standard text, denoted as B4; then the sentence-level text evaluation result can be denoted as: B 译 = (B3 + B4) / 2;

[0119] Further, the word-level evaluation result of the text in the target language can be denoted as (A 语 + A 译 ) / 2; the word-level evaluation result of the text in the target language can be denoted as (B 语 + B 译 ) / 2; the translation quality of the text in the target language can be denoted as ((A 语 + A 译 ) / 2 + (B 语 + B 译 ) / 2) / 2.

[0120] The translation quality evaluation device provided by the present invention is described below. The translation quality evaluation device described below can be correspondingly referred to the translation quality evaluation method described above.

[0121] Figure 7 is a structural schematic diagram of the translation quality evaluation device provided by the present invention, asFigure 7 As shown, the device includes:

[0122] A voice and text determination unit 710, configured to determine the source language voice and the target language text;

[0123] A voice evaluation unit 720, configured to perform source language standard translation on the source language voice to obtain a source language standard text, and determine a voice evaluation result based on the source language standard text;

[0124] A text evaluation unit 730, configured to determine a text evaluation result based on the target language text;

[0125] A translation quality determination unit 740, configured to determine the translation quality of the target language text based on the voice evaluation result and the text evaluation result.

[0126] The translation quality evaluation device provided by the embodiment of the present invention performs source language standard translation on the source language voice to obtain a source language standard text, determines a voice evaluation result based on the source language standard text; and determines a text evaluation result based on the target language text; and then determines the translation quality of the target language text based on the voice evaluation result and the text evaluation result, improving the accuracy of translation quality evaluation.

[0127] Based on any of the above embodiments, the voice evaluation unit 720 is further configured to:

[0128] Determine a word-level voice evaluation result based on the translation accuracy rate of the word segmentation in the source language standard text for the source language voice;

[0129] And / or, determine a sentence-level voice evaluation result based on the semantic accuracy and / or expression fluency of the sentence segmentation in the source language standard text;

[0130] Determine a voice evaluation result based on the word-level voice evaluation result and / or the sentence-level voice evaluation result.

[0131] Based on any of the above embodiments, the translation quality evaluation device further includes a translation accuracy rate determination unit, configured to:

[0132] Determine a first technical field based on the technical field of the word segmentation in the transcription text of the source language voice;

[0133] Determine the translation accuracy rate of the word segmentation in the source language standard text for the source language voice based on the first technical field.

[0134] Based on any of the above embodiments, the text evaluation unit 730 is further configured to:

[0135] Determine the word-level text evaluation result based on the translation accuracy of the word segmentation in the target language text for the source language standard text;

[0136] And / or, perform source language standard translation on the target language text to obtain a translation standard text, and determine the sentence-level text evaluation result based on the semantic accuracy and / or expression fluency of the clauses in the translation standard text;

[0137] Determine the text evaluation result based on the word-level text evaluation result and / or the sentence-level text evaluation result.

[0138] Based on any of the above embodiments, the translation accuracy determination unit is further configured to:

[0139] Determine a second technical field based on the technical field of the word segmentation in the target language text;

[0140] Determine the translation accuracy of the word segmentation in the target language text for the source language standard text based on the second technical field.

[0141] Based on any of the above embodiments, the translation quality evaluation device further includes a semantic accuracy determination unit, configured to:

[0142] Determine a target standard word order, where the target standard word order is determined based on the standard word order of the source language under the first technical field and / or the standard word order of the target language under the second technical field;

[0143] Based on the target standard word order, adjust the word order of the clauses in the source language standard text or the translation standard text, and obtain the semantic accuracy of the clauses in the source language standard text or the translation standard text after the word order adjustment.

[0144] Based on any of the above embodiments, the translation quality evaluation device further includes an expression fluency determination unit, configured to:

[0145] Determine the expression fluency of the clauses in the source language standard text or the translation standard text based on the relevance of the word segmentation in the clauses of the source language standard text or the translation standard text.

[0146] Figure 8 Illustrates a schematic diagram of the physical structure of an electronic device, such as Figure 8As shown in the figure, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete communication with each other through the communication bus 840. The processor 810 may call the logical instructions in the memory 830 to execute a translation quality evaluation method, which includes: determining the source language speech and the target language text; performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text; determining a text evaluation result based on the target language text; and determining the translation quality of the target language text based on the speech evaluation result and the text evaluation result.

[0147] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as an independent product, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0148] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the translation quality evaluation method provided by the above-mentioned various methods, which includes: determining the source language speech and the target language text; performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text; determining a text evaluation result based on the target language text; and determining the translation quality of the target language text based on the speech evaluation result and the text evaluation result.

[0149] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the translation quality evaluation method provided by the above-mentioned various methods. The method includes: determining a source language speech and a target language text; performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text; determining a text evaluation result based on the target language text; and determining the translation quality of the target language text based on the speech evaluation result and the text evaluation result.

[0150] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0151] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements on some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for evaluating translation quality, characterized in that, including: determining the source language speech and the target language text; performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text; determining a text evaluation result based on the target language text; determining the translation quality of the target language text based on the speech evaluation result and the text evaluation result; The determining the speech evaluation result based on the source language standard text includes: determining a word-level speech evaluation result based on the translation accuracy rate of the word segmentation in the source language standard text for the source language speech; and / or, determining a sentence-level speech evaluation result based on the semantic accuracy and / or expression fluency of the clauses in the source language standard text; determining the speech evaluation result based on the word-level speech evaluation result and / or the sentence-level speech evaluation result.

2. The translation quality evaluation method according to claim 1, characterized in that The translation accuracy rate of the word segmentation in the source language standard text for the source language speech is determined based on the following steps: determining a first technical field based on the technical field of the word segmentation in the transcription text of the source language speech; determining the translation accuracy rate of the word segmentation in the source language standard text for the source language speech based on the first technical field.

3. The translation quality evaluation method according to claim 1, characterized in that The determining the text evaluation result based on the target language text includes: determining a word-level text evaluation result based on the translation accuracy rate of the word segmentation in the target language text for the source language standard text; and / or, performing source language standard translation on the target language text to obtain a translation standard text, and determining a sentence-level text evaluation result based on the semantic accuracy and / or expression fluency of the clauses in the translation standard text; determining the text evaluation result based on the word-level text evaluation result and / or the sentence-level text evaluation result.

4. The translation quality evaluation method according to claim 3, wherein The translation accuracy rate of the word segmentation in the target language text for the source language standard text is determined based on the following steps: determining a second technical field based on the technical field of the word segmentation in the target language text; determining the translation accuracy rate of the word segmentation in the target language text for the source language standard text based on the second technical field.

5. The translation quality evaluation method according to any one of claims 3 or 4, characterized in that The semantic accuracy of the clause is determined based on the following steps: determining a target standard word order, where the target standard word order is determined based on the standard word order of the source language under the first technical field and / or the standard word order of the target language under the second technical field; adjusting the word order of the clause in the source language standard text or the translation standard text based on the target standard word order, and obtaining the semantic accuracy of the clause in the source language standard text or the translation standard text after the word order adjustment.

6. The translation quality evaluation method according to any one of claims 3 or 4, characterized in that, The expression fluency of the clause is determined based on the following steps: determining the expression fluency of the clause in the source language standard text or the translation standard text based on the relevance of the word segmentation in the clause of the source language standard text or the translation standard text.

7. A translation quality evaluation device, characterized in that, including: a speech and text determination unit for determining the source language speech and the target language text; a speech evaluation unit for performing source language standard translation on the source language speech to obtain a source language standard text, and determining a speech evaluation result based on the source language standard text; A text evaluation unit for determining a text evaluation result based on the target language text; A translation quality determination unit for determining the translation quality of the target language text based on the speech evaluation result and the text evaluation result; The speech evaluation unit is configured to: Determine a word-level speech evaluation result based on the translation accuracy rate of the word segmentation in the source language standard text for the source language speech; And / or determine a sentence-level speech evaluation result based on the semantic accuracy and / or expression fluency of the sentence clauses in the source language standard text; Determine a speech evaluation result based on the word-level speech evaluation result and / or the sentence-level speech evaluation result.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the translation quality evaluation method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the translation quality evaluation method according to any one of claims 1 to 6.

Images