Simultaneous interpretation intelligent proofreading method and device and storage medium

Abstract

The invention relates to the field of intelligent correction, and particularly discloses an intelligent proofreading method and device for simultaneous interpretation and a storage medium. And semantic coding and semantic association based on word granularity and sentence granularity are carried out on the target language text by utilizing a text semantic analysis and processing algorithm based on artificial intelligence, so that correction suggestions are intelligently obtained. Through the mode, the system can analyze the translation of the simultaneous interpreter in real time, identify and suggest to modify grammar errors or inaccurate expressions in translation by utilizing intelligent correction, quickly provide correction suggestions, help the interpreter to improve in time, reduce the dependence on manual proofreading and reduce the labor cost of translation items.

Description

Technical Field

[0001] The present application relates to the field of intelligent correction, and more specifically, to a simultaneous interpretation intelligent proofreading method, device and storage medium. Background Art

[0002] Simultaneous interpretation, also known as simultaneous translation or simultaneous interpretation, is a method of translation in which the translator translates the speech into the target language almost simultaneously without interrupting the speaker. Proofreading simultaneous interpretation is a meticulous and demanding process that is essential to ensure the high quality of translation services. Proofreading simultaneous interpretation ensures that the translation is faithful to the original meaning and free of misunderstandings or errors.

[0003] However, since traditional proofreading is mostly done manually, the proofreading speed is slow, especially in large meetings or long speeches, and it is difficult to keep up with the speed of simultaneous interpretation. At the same time, there may be differences in the understanding and judgment of different proofreaders, and long proofreading work will cause proofreader fatigue, thus affecting the accuracy and consistency of proofreading.

[0004] Therefore, an optimized intelligent proofreading method for simultaneous interpretation is desired. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiments of the present application provide a method, device and storage medium for intelligent proofreading of simultaneous interpretation, which obtains the target language text by performing speech recognition on the speech translated by the simultaneous interpreter, and uses the text semantic analysis and processing algorithm based on artificial intelligence to perform semantic encoding and semantic association based on word granularity and sentence granularity on the target language text, so as to intelligently obtain correction suggestions. In this way, the system can analyze the translation of the simultaneous interpreter in real time, use intelligent correction to identify and suggest modifications to grammatical errors or inaccurate expressions in the translation, and quickly provide correction suggestions to help the translator make timely improvements, reduce the reliance on manual proofreading, and reduce the manpower cost of the translation project.

[0006] According to one aspect of the present application, there is provided a method for intelligent proofreading of simultaneous interpretation, which comprises:

[0007] Perform speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text;

[0008] Performing word segmentation and sentence segmentation processing on the target language text respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text;

[0009] Performing text semantic encoding on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors;

[0010] Inputting the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors into a text semantic context associator respectively, so as to obtain the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector;

[0011] Fusion of the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector to obtain a target language text multi-scale global semantic association coding vector;

[0012] Based on the multi-scale global semantic association encoding vector of the target language text, correction suggestions are obtained.

[0013] According to another aspect of the present application, there is provided an intelligent proofreading device for simultaneous interpretation, comprising:

[0014] The translation speech recognition module is used to perform speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text;

[0015] A target language text word segmentation and sentence segmentation processing module is used to perform word segmentation and sentence segmentation processing on the target language text to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text;

[0016] A target language word-granularity text sentence-granularity semantic encoding module is used to perform text semantic encoding on the sequence of the target language word-granularity text and the sequence of the target language sentence-granularity text to obtain a sequence of target language word-granularity text semantic encoding vectors and a sequence of target language sentence-granularity text semantic encoding vectors;

[0017] A target language semantic association encoding module, used for inputting the sequence of the target language word granularity text semantic encoding vectors and the sequence of the target language sentence granularity text semantic encoding vectors into a text semantic context associator respectively, so as to obtain a target language word granularity semantic association encoding vector and a target language sentence granularity semantic association encoding vector;

[0018] A target language text multi-scale fusion module, used to fuse the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector to obtain a target language text multi-scale global semantic association coding vector;

[0019] The correction suggestion generation module is used to obtain correction suggestions based on the multi-scale global semantic association encoding vector of the target language text.

[0020] According to another aspect of the present application, a storage medium is also provided, comprising a computer program, which, when executed on a computer device, enables a processing unit in the computer device to execute the intelligent proofreading method for simultaneous interpretation as described above.

[0021] Compared with the prior art, the present application provides a method, device and storage medium for intelligent proofreading of simultaneous interpretation, which obtains the target language text by performing speech recognition on the speech translated by the simultaneous interpreter, and uses the text semantic analysis and processing algorithm based on artificial intelligence to perform semantic encoding and semantic association on the target language text based on word granularity and sentence granularity, so as to intelligently obtain correction suggestions. In this way, the system can analyze the translation of the simultaneous interpreter in real time, use intelligent correction to identify and suggest modifications to grammatical errors or inaccurate expressions in the translation, and quickly provide correction suggestions to help the translator make timely improvements, reduce dependence on manual proofreading, and reduce the labor cost of the translation project. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 The figure is a flow chart of the intelligent proofreading method of simultaneous interpretation according to an embodiment of the present application.

[0024] Figure 2 Schematic diagram of the architecture of the intelligent proofreading method for simultaneous interpretation according to an embodiment of the present application.

[0025] Figure 3 The present invention is a flowchart of performing text semantic encoding on the sequence of target language word-granularity text and the sequence of target language sentence-granularity text in the intelligent proofreading method for simultaneous interpretation according to an embodiment of the present application to obtain a sequence of target language word-granularity text semantic encoding vectors and a sequence of target language sentence-granularity text semantic encoding vectors.

[0026] Figure 4 4 is a block diagram of an intelligent proofreading device for simultaneous interpretation according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0028] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0029] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0030] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0031] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0032] Simultaneous interpretation, also known as simultaneous translation or simultaneous interpretation, is a method of translation in which the translator translates the speech into the target language almost simultaneously without interrupting the speaker. Proofreading simultaneous interpretation is a meticulous and demanding process that is essential to ensure the high quality of translation services. Proofreading simultaneous interpretation ensures that the translation is faithful to the original meaning and free of misunderstandings or errors.

[0033] However, since traditional proofreading is mostly done manually, the proofreading speed is slow, especially in large meetings or long speeches, and it is difficult to keep up with the speed of simultaneous interpretation. At the same time, there may be differences in the understanding and judgment of different proofreaders, and long proofreading work will cause proofreader fatigue, thus affecting the accuracy and consistency of proofreading.

[0034] Therefore, a simultaneous interpretation intelligent proofreading method is desired, which obtains the target language text by performing speech recognition on the speech translated by the simultaneous interpreter, and uses the text semantic analysis and processing algorithm based on artificial intelligence to perform semantic encoding and semantic association on the target language text based on word granularity and sentence granularity, so as to intelligently obtain correction suggestions. In this way, the system can analyze the translation of the simultaneous interpreter in real time, use intelligent correction to identify and suggest modifications to grammatical errors or inaccurate expressions in the translation, and quickly provide correction suggestions to help the translator make timely improvements, reduce the reliance on manual proofreading, and reduce the labor cost of the translation project.

[0035] Figure 1 The figure is a flow chart of the intelligent proofreading method of simultaneous interpretation according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of the intelligent proofreading method for simultaneous interpretation according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the intelligent proofreading method for simultaneous interpretation according to the embodiment of the present application includes: S110, performing speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text; S120, performing word segmentation and sentence segmentation processing on the target language text respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text; S130, performing text semantic encoding on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors; S140, input the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors into the text semantic context associator respectively to obtain the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector; S150, fuse the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector to obtain the target language text multi-scale global semantic association coding vector; S160, obtain correction suggestions based on the target language text multi-scale global semantic association coding vector.

[0036] In step S110, the speech translated by the simultaneous interpreter is subjected to speech recognition to obtain the target language text. It should be understood that considering that the speech translated by the simultaneous interpreter is subjected to speech recognition to obtain the target language text, it is one of the key steps to realize the intelligent proofreading method of simultaneous interpretation. Specifically, speech recognition technology can automatically convert speech into text, and provide automated text input for subsequent semantic analysis and correction. Based on this, in order to be able to more accurately and meticulously analyze the content translated by the simultaneous interpreter, so as to perform more accurate proofreading, in the technical scheme of the present application, the speech translated by the simultaneous interpreter is subjected to speech recognition to obtain the target language text. That is, through speech recognition, the intelligent proofreading method of simultaneous interpretation can automatically process the translation content, and provide technical support for realizing fast and accurate translation correction, which is very important for improving the overall quality and efficiency of simultaneous interpretation.

[0037] In step S120, the target language text is subjected to word segmentation and sentence segmentation processing respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text. Accordingly, considering that the target language text is a long text composed of multiple words or sentences, these words or sentences have an important role and influence on understanding and analyzing the content and semantic information of the target text. And the target language text contains semantic information about the target language text between different scales. Therefore, in order to more carefully analyze and process the semantic information of different scales in the target language text, in the technical solution of the present application, the target language text is subjected to word segmentation and sentence segmentation processing respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text. That is to say, the target language text is subjected to word segmentation and sentence segmentation processing, which helps to identify the language structure in the target language text, such as distinguishing between single words and complete sentences, so as to better understand the text. In addition, word segmentation and sentence segmentation are the basis of grammatical analysis, which helps to determine the parts of speech, syntactic structure and semantic relations, making it easier to extract key information such as entities, phrases and concepts, and can provide structured text data for intelligent proofreading methods of simultaneous interpretation, thereby achieving more effective semantic encoding and semantic association analysis, and improving the accuracy and quality of translation.

[0038] In step S130, text semantic encoding is performed on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors. Figure 3The present invention is a flowchart of performing text semantic encoding on the target language word-granularity text sequence and the target language sentence-granularity text sequence in the intelligent proofreading method of simultaneous interpretation according to the embodiment of the present application to obtain a sequence of target language word-granularity text semantic encoding vectors and a sequence of target language sentence-granularity text semantic encoding vectors. Specifically, in the embodiment of the present application, Figure 3 As shown, text semantic encoding is performed on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors, including: S210, inputting the sequence of target language word granularity text into a target language text semantic word encoder to obtain a sequence of target language word granularity text semantic encoding vectors; S220, inputting the sequence of target language sentence granularity text into a target language text semantic sentence encoder to obtain a sequence of target language sentence granularity text semantic encoding vectors.

[0039] Specifically, in step S210, the sequence of the target language word granularity text is input into the target language text semantic word encoder to obtain the sequence of the target language word granularity text semantic encoding vectors. Specifically, in an embodiment of the present application, the sequence of the target language word granularity text is input into the target language text semantic word encoder to obtain the sequence of the target language word granularity text semantic encoding vectors, including: inputting the sequence of the target language word granularity text into the target language text semantic word encoder containing a word embedding layer to obtain the sequence of the target language word granularity text semantic encoding vectors. Accordingly, considering that there is a semantic association between the mutual contexts between the individual target language word granularity texts in the sequence of the target language word granularity text. The word embedding layer can convert vocabulary into a dense vector form, and these vectors capture the semantic information of the word. Therefore, in the technical solution of the present application, the sequence of the target language word granularity text is input into the target language text semantic word encoder containing a word embedding layer to obtain the sequence of the target language word granularity text semantic encoding vectors. That is to say, through word embedding, the model can better understand the relationship between words in the target language text, including synonyms, antonyms, etc. In particular, the word embedding usually takes into account the context of the word in the corpus, and the generated vector can reflect the usage of the word in different contexts. In this way, the target language word-granular text data can be effectively converted into a sequence of target language word-granular text semantic encoding vectors that can capture rich semantic information, providing a basis for subsequent semantic encoding and semantic association analysis, thereby improving the quality of intelligent proofreading of simultaneous interpretation.

[0040] In step S220, the sequence of the target language sentence granularity text is input into the target language text semantic sentence encoder to obtain the sequence of the target language sentence granularity text semantic coding vector. Specifically, in an embodiment of the present application, the sequence of the target language sentence granularity text is input into the target language text semantic sentence encoder to obtain the sequence of the target language sentence granularity text semantic coding vector, including: the sequence of the target language sentence granularity text is input into the target language text semantic sentence encoder containing a sentence embedding layer to obtain the sequence of the target language sentence granularity text semantic coding vector. Similarly, considering that there is also a semantic association relationship between the contexts between the various target language sentence granularity texts in the sequence of the target language sentence granularity text, and considering that the sentence embedding layer can capture the semantic information of the entire sentence, not just a single word, which is crucial to understanding the overall meaning of the sentence. Therefore, in the technical solution of the present application, the sequence of the target language sentence granularity text is input into the target language text semantic sentence encoder containing a sentence embedding layer to obtain the sequence of the target language sentence granularity text semantic coding vector. It should be understood that the lexical contextual relationship in each target language sentence-granular text in the sequence of the target language sentence-granular text is very important for understanding the meaning of the sentence, and the sentence embedding layer can better model these contextual relationships. In particular, in the target language sentence-granular text, important semantic information may span multiple words, and the sentence embedding layer helps to capture such long-distance dependencies and effectively convert sentences into sequences of target language sentence-granular text semantic encoding vectors that can capture rich semantic information.

[0041] In step S140, the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors are respectively input into a text semantic context associator to obtain a target language word granularity semantic association coding vector and a target language sentence granularity semantic association coding vector. Specifically, in an embodiment of the present application, the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors are respectively input into a text semantic context associator to obtain a target language word granularity semantic association coding vector and a target language sentence granularity semantic association coding vector, including: the sequence of the target language word granularity text semantic coding vector and the sequence of the target language sentence granularity text semantic coding vector are respectively input into a text semantic context associator based on a transformer mechanism to obtain the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector. Accordingly, considering that the sequence of the target language word-granularity text semantic coding vectors and the sequence of the target language sentence-granularity text semantic coding vectors have different degrees of semantic association relationships at different time spans, the Transformer model can capture the long-distance dependencies between different semantic spans in the sequence through the self-attention mechanism, which helps to understand the long-distance semantic connections in the text. Based on this, in the technical solution of the present application, the sequence of the target language word-granularity text semantic coding vectors and the sequence of the target language sentence-granularity text semantic coding vectors are respectively input into the text semantic context associator based on the transformer mechanism to respectively capture and mine the language association dependencies between each different word granularity and granularity semantic span, and obtain the target language word-granularity semantic association coding vector and the target language sentence-granularity semantic association coding vector.

[0042] In step S150, the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector are fused to obtain a multi-scale global semantic association coding vector of the target language text. Specifically, in an embodiment of the present application, the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector are fused to obtain a multi-scale global semantic association coding vector of the target language text, including: based on the target language word granularity semantic association coding vector, the target language sentence granularity semantic association coding vector is enhanced by a perspective metric based on a multi-perspective feature relationship metric to obtain an optimized target language sentence granularity semantic association coding vector; the optimized target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector are fused to obtain a multi-scale global semantic association coding vector of the target language text.

[0043] Specifically, based on the target language word-granularity semantic association coding vector, the target language sentence-granularity semantic association coding vector is enhanced by perspective measurement based on multi-perspective feature relationship measurement to obtain an optimized target language sentence-granularity semantic association coding vector. In particular, in the technical solution of the present application, it is considered that the semantic coding vector of word granularity focuses on the semantic information of a single word, while the semantic coding vector of sentence granularity focuses on the semantic information at the sentence level. This difference in granularity may result in the inability to fully reflect the semantic information at the sentence level in the semantic coding of word granularity during fusion, resulting in a lack of causal association. The semantics of a sentence often depends on contextual information, while the semantics of a word are relatively independent. During the processing, if the contextual information of the sentence is not fully captured, the sentence-granularity semantic coding vector may not provide sufficient prior information during fusion. The semantic encoder of the target language text may be designed to focus more on the extraction of semantic information at word granularity, while the extraction of semantic information at sentence granularity is not sufficient. This design deviation may result in the sentence-granularity semantic coding vector being unable to form an effective causal association with the word-granularity semantic coding vector when fused, which in turn results in the target language sentence-granularity semantic association coding vector having a missing a posteriori-priori causal association relative to the target language word-granularity semantic association coding vector. In order to solve this problem, in the technical solution of the present application, based on the target language word-granularity semantic association coding vector, the target language sentence-granularity semantic association coding vector is enhanced by a perspective metric based on a multi-perspective feature relationship metric to obtain an optimized target language sentence-granularity semantic association coding vector.

[0044] More specifically, in an embodiment of the present application, based on the target language word granularity semantic association coding vector, the target language sentence granularity semantic association coding vector is subjected to perspective metric enhancement based on multi-perspective feature relationship metric to obtain an optimized target language sentence granularity semantic association coding vector, including: calculating the positional difference between the target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector to obtain a first perspective metric representation vector; calculating the positional dot product between the target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector to obtain a second perspective metric representation vector; calculating the positional dot product between the target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector The first viewpoint metric representation vector is summed up by position to obtain a third viewpoint metric representation vector; the gradient amplitude of each position in the first viewpoint metric representation vector, the second viewpoint metric representation vector and the third viewpoint metric representation vector is calculated to obtain a first viewpoint metric perception vector, a second viewpoint metric perception vector and a third viewpoint metric perception vector; the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector are input into a backward machine learning reverse multi-viewpoint perception optimization module to obtain a reverse multi-viewpoint perception optimization representation vector; the positional dot product between the reverse multi-viewpoint perception optimization representation vector and the target language sentence granularity semantic association encoding vector is calculated to obtain the optimized target language sentence granularity semantic association encoding vector.

[0045] The backward machine learning reverse multi-view perception optimization module is expressed as follows:

[0046]

[0047] in, and represents the eigenvalues ​​of the i-th position of the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector, respectively, and μ represents the three eigenvalues ​​of the corresponding positions in the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector , σ is the standard deviation of the three eigenvalues ​​at corresponding positions in the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector, represents the maximum value of the three eigenvalues ​​at the corresponding positions in the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector, ReLU represents a linear rectification function, softmax represents a normalized exponential function, and v ri Represents the eigenvalue of the i-th position of the inverse multi-view perceptual optimization representation vector.

[0048] Specifically, in the technical solution of the present application, based on the target language word granularity semantic association coding vector, the target language sentence granularity semantic association coding vector is subjected to perspective metric enhancement optimization based on multi-view feature relationship metric, and the process includes firstly calculating the position difference, dot multiplication and sum between the target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector through multi-view analysis (Multi-View Analysis) to obtain three perspective metric representation vectors, and quantify the relationship between features from different angles. Then, the gradient amplitude of each position in these perspective metric representation vectors is calculated through backward error perception to identify key change points and provide a basis for feature optimization.

[0049] Then, the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector are input into a backward machine learning reverse multi-view perception optimization module, wherein the backward machine learning reverse multi-view perception optimization module adopts a squeeze-incentive mechanism to capture the significant comprehensive ability factor of the feature importance perception of the feature sub-dimension distribution of the corresponding positions in the first to third viewpoint metric perception vectors. Then, the positional dot product between the reverse multi-view perception optimization representation vector and the target language sentence granularity semantic association encoding vector is calculated, so as to perform backward optimization perception on the target language sentence granularity semantic association encoding vector, so as to adjust the target language sentence granularity semantic association encoding vector, so as to better associate it with the target language word granularity semantic association encoding vector, and enhance the causal relationship between the target language sentence granularity semantic association encoding vector and the target language word granularity semantic association encoding vector.

[0050] Specifically, the optimized target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector are fused to obtain the multi-scale global semantic association coding vector of the target language text. Accordingly, considering that the optimized target language sentence granularity semantic association coding vector expresses the semantic information of the entire target language sentence after feature optimization, it is helpful to understand the meaning of the sentence and the context association. The target language word granularity semantic association coding vector represents the semantic information about the words or phrases of the target language, which helps to understand the semantic association and meaning between words. Based on this, in order to more comprehensively capture the semantic features of the target language text and improve the understanding and representation capabilities of the text semantics, in the technical solution of the present application, the optimized target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector are fused to obtain the multi-scale global semantic association coding vector of the target language text. That is, by fusing the optimized target language sentence-granularity semantic association encoding vector and the target language word-granularity semantic association encoding vector, the semantic association of the entire target language text can be more comprehensively understood while retaining the word-level detail information and sentence-level context information, thereby obtaining a comprehensive multi-scale global semantic association encoding vector.

[0051] In step S160, based on the multi-scale global semantic association coding vector of the target language text, correction suggestions are obtained. Specifically, in an embodiment of the present application, based on the multi-scale global semantic association coding vector of the target language text, correction suggestions are obtained, including: the multi-scale global semantic association coding of the target language text is input into a generator to generate the correction suggestions. That is, by generating and processing the multi-scale global semantic association coding of the target language text, correction suggestions are intelligently obtained. In this way, the system can analyze the translation of the simultaneous interpreter in real time, use intelligent correction to identify and suggest modifications to grammatical errors or inaccurate expressions in the translation, and quickly provide correction suggestions to help interpreters improve in time, reduce dependence on manual proofreading, and reduce the labor cost of translation projects.

[0052] In summary, the intelligent proofreading method for simultaneous interpretation based on the embodiment of the present application is explained, which obtains the target language text by performing speech recognition on the speech translated by the simultaneous interpreter, and uses the text semantic analysis and processing algorithm based on artificial intelligence to perform semantic encoding and semantic association based on word granularity and sentence granularity on the target language text, thereby intelligently obtaining correction suggestions. In this way, the system can analyze the translation of the simultaneous interpreter in real time, use intelligent correction to identify and suggest modifications to grammatical errors or inaccurate expressions in the translation, and quickly provide correction suggestions to help the translator make timely improvements, reduce dependence on manual proofreading, and reduce the labor cost of the translation project.

[0053] Figure 4 FIG. 1 is a block diagram of an intelligent proofreading device for simultaneous interpretation according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the intelligent proofreading device 100 for simultaneous interpretation includes: a translation speech recognition module 110, which is used to perform speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text; a target language text word segmentation and sentence processing module 120, which is used to perform word segmentation and sentence processing on the target language text respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text; a target language word granularity text sentence granularity semantic encoding module 130, which is used to perform text semantic encoding on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors. a target language semantic association encoding module 140, used to input the sequence of target language word granularity text semantic encoding vectors and the sequence of target language sentence granularity text semantic encoding vectors into a text semantic context associator respectively, so as to obtain a target language word granularity semantic association encoding vector and a target language sentence granularity semantic association encoding vector; a target language text multi-scale fusion module 150, used to fuse the target language word granularity semantic association encoding vector and the target language sentence granularity semantic association encoding vector, so as to obtain a target language text multi-scale global semantic association encoding vector; a correction suggestion generation module 160, used to obtain a correction suggestion based on the target language text multi-scale global semantic association encoding vector.

[0054] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned intelligent proofreading device for simultaneous interpretation have been described in the above reference. Figures 1 to 3 The invention has been introduced in detail in the description of the intelligent proofreading method for simultaneous interpretation, and therefore, its repeated description will be omitted.

[0055] The embodiment of the present invention further relates to a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the above method embodiment is implemented.

[0056] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0057] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent proofreading method for simultaneous interpretation, characterized in that: include: Perform speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text; Performing word segmentation and sentence segmentation processing on the target language text respectively to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text; Performing text semantic encoding on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors; Inputting the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors into a text semantic context associator respectively, so as to obtain the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector; Fusion of the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector to obtain a target language text multi-scale global semantic association coding vector; Based on the multi-scale global semantic association encoding vector of the target language text, correction suggestions are obtained.

2. The intelligent proofreading method for simultaneous interpretation according to claim 1, characterized in that: Performing text semantic encoding on the sequence of target language word granularity text and the sequence of target language sentence granularity text to obtain a sequence of target language word granularity text semantic encoding vectors and a sequence of target language sentence granularity text semantic encoding vectors, including: Inputting the target language word granularity text sequence into the target language text semantic word encoder to obtain the target language word granularity text semantic encoding vector sequence; The sequence of the target language sentence granularity text is input into the target language text semantic sentence encoder to obtain a sequence of the target language sentence granularity text semantic encoding vectors.

3. The intelligent proofreading method for simultaneous interpretation according to claim 2, characterized in that: Inputting the target language word granularity text sequence into the target language text semantic word encoder to obtain the target language word granularity text semantic encoding vector sequence, including: The sequence of the target language word granularity text is input into a target language text semantic word encoder including a word embedding layer to obtain a sequence of the target language word granularity text semantic encoding vectors.

4. The intelligent proofreading method for simultaneous interpretation according to claim 3, characterized in that: Inputting the target language sentence granularity text sequence into the target language text semantic sentence encoder to obtain the target language sentence granularity text semantic encoding vector sequence, including: The sequence of the target language sentence granularity text is input into a target language text semantic sentence encoder including a sentence embedding layer to obtain a sequence of the target language sentence granularity text semantic encoding vectors.

5. The intelligent proofreading method for simultaneous interpretation according to claim 4, characterized in that: Inputting the sequence of the target language word granularity text semantic coding vectors and the sequence of the target language sentence granularity text semantic coding vectors into a text semantic context associator respectively to obtain a target language word granularity semantic association coding vector and a target language sentence granularity semantic association coding vector, including: The sequence of the target language word-granularity text semantic coding vectors and the sequence of the target language sentence-granularity text semantic coding vectors are respectively input into the text semantic context associator based on the transformer mechanism to obtain the target language word-granularity semantic association coding vector and the target language sentence-granularity semantic association coding vector.

6. The intelligent proofreading method for simultaneous interpretation according to claim 5, characterized in that: The target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector are integrated to obtain a target language text multi-scale global semantic association coding vector, including: Based on the target language word granularity semantic association coding vector, the target language sentence granularity semantic association coding vector is enhanced by perspective metric based on multi-perspective feature relationship metric to obtain an optimized target language sentence granularity semantic association coding vector; The optimized target language sentence granularity semantic association coding vector and the target language word granularity semantic association coding vector are fused to obtain the target language text multi-scale global semantic association coding vector.

7. The intelligent proofreading method for simultaneous interpretation according to claim 6, characterized in that: Based on the target language word granularity semantic association coding vector, the target language sentence granularity semantic association coding vector is enhanced by perspective metric based on multi-perspective feature relationship metric to obtain an optimized target language sentence granularity semantic association coding vector, including: Calculating the position difference between the target language sentence granularity semantic association encoding vector and the target language word granularity semantic association encoding vector to obtain a first viewpoint metric representation vector; Calculating the positional dot product between the target language sentence granularity semantic association encoding vector and the target language word granularity semantic association encoding vector to obtain a second viewpoint metric representation vector; Calculating the positional sum of the target language sentence granularity semantic association encoding vector and the target language word granularity semantic association encoding vector to obtain a third viewpoint metric representation vector; Calculating the gradient magnitude of each position in the first viewpoint metric representation vector, the second viewpoint metric representation vector, and the third viewpoint metric representation vector to obtain a first viewpoint metric perception vector, a second viewpoint metric perception vector, and a third viewpoint metric perception vector; Inputting the first viewpoint metric perception vector, the second viewpoint metric perception vector and the third viewpoint metric perception vector into a backward machine learning inverse multi-viewpoint perception optimization module to obtain a reverse multi-viewpoint perception optimization representation vector; The positional dot product between the reverse multi-view perception optimization representation vector and the target language sentence granularity semantic association encoding vector is calculated to obtain the optimized target language sentence granularity semantic association encoding vector.

8. The intelligent proofreading method for simultaneous interpretation according to claim 7, characterized in that: Based on the multi-scale global semantic association encoding vector of the target language text, correction suggestions are obtained, including: The multi-scale global semantic association encoding of the target language text is input into a generator to generate the correction suggestion.

9. An intelligent proofreading device for simultaneous interpretation, characterized in that: include: The translation speech recognition module is used to perform speech recognition on the speech translated by the simultaneous interpreter to obtain the target language text; A target language text word segmentation and sentence segmentation processing module is used to perform word segmentation and sentence segmentation processing on the target language text to obtain a sequence of target language word granularity text and a sequence of target language sentence granularity text; A target language word-granularity text sentence-granularity semantic encoding module is used to perform text semantic encoding on the sequence of the target language word-granularity text and the sequence of the target language sentence-granularity text to obtain a sequence of target language word-granularity text semantic encoding vectors and a sequence of target language sentence-granularity text semantic encoding vectors; A target language semantic association encoding module, used for inputting the sequence of the target language word granularity text semantic encoding vectors and the sequence of the target language sentence granularity text semantic encoding vectors into a text semantic context associator respectively, so as to obtain a target language word granularity semantic association encoding vector and a target language sentence granularity semantic association encoding vector; A target language text multi-scale fusion module, used to fuse the target language word granularity semantic association coding vector and the target language sentence granularity semantic association coding vector to obtain a target language text multi-scale global semantic association coding vector; The correction suggestion generation module is used to obtain correction suggestions based on the multi-scale global semantic association encoding vector of the target language text.

10. A storage medium, characterized in that , including a computer program, which, when executed on a computer device, enables a processing unit in the computer device to execute the intelligent proofreading method for simultaneous interpretation as described in any one of claims 1 to 8.

Images