Machine-assisted English-Chinese technology translation error analysis system and method based on SVM (Support Vector Machine)

Abstract

The invention relates to the technical field of machine translation, and discloses a machine-assisted English-Chinese technology translation error analysis system and method based on an SVM (Support Vector Machine), and the method comprises the steps that a corpus construction module constructs a professional technology text corpus, and translates an obtained Chinese machine translation text; the error annotation module carries out error annotation on the translated text based on an annotation rule; the feature extraction module is used for extracting key features from the Chinese translated text and the translated text; the model building module converts the key features and the error labeling information into feature vectors and divides the feature vectors into a training set and a test set; training a translation error detection model; and the error detection module is used for inputting the to-be-analyzed machine translation text into the trained translation error detection model to obtain an error result. According to the method, an accurate classification model can be constructed by learning a large amount of sample data. Errors in a machine translation text are accurately recognized, classified and labeled, manual proofreading is assisted, and then the overall quality of English-Chinese technical translation is improved.

Claims

1. A machine-assisted English-Chinese technical translation error analysis system based on SVM, characterized in that: include: A corpus construction module is used to construct a professional technical text corpus, wherein the professional technical text corpus includes sentence-aligned English original texts and Chinese translations, and a Chinese machine translation text obtained by translating the English original texts using a machine translation engine; An error marking module, used to establish marking rules, and mark errors of the Chinese machine translation text based on the marking rules, wherein the error marking information includes error type, error location and correct translation; A feature extraction module, used to preprocess the Chinese translation and the Chinese machine translation text, and extract key features from the preprocessed Chinese translation and the Chinese machine translation text, wherein the key features include lexical features, grammatical features and semantic features; A model building module, for converting the key features and error annotation information into feature vectors, and dividing the feature vectors into a training set and a test set; using a support vector machine as a classifier, training a translation error detection model according to the training set and the test set, and adjusting the parameters of the translation error detection model using a cross-validation method; The error detection module is used to input the machine translation text to be analyzed into the trained translation error detection model to obtain the error result of the machine translation text.

2. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1 is characterized in that: The professional technical text corpus includes: professional technical literature, patent documents and product manuals.

3. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1 is characterized in that: When the error marking module establishes marking rules, the marking rules include: marking of term translation errors, marking of misjudgment of part of speech, marking of improper word order, marking of grammatical errors and marking of semantic understanding deviations.

4. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1, characterized in that: When the feature extraction module preprocesses the Chinese translation and the Chinese machine translation text, the preprocessing includes word segmentation processing and stop word removal processing.

5. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1, characterized in that: When the feature extraction module extracts key features from the preprocessed Chinese translation and Chinese machine translation text, the lexical features include word frequency, part of speech, stem, and popularity of words in professional fields; The grammatical features include: sentence components, sentence patterns, tense, voice, clause types and nesting; The semantic features include: semantic similarity between words and semantic coherence of sentences.

6. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1, characterized in that: When the model building module divides the feature vector into a training set and a test set, it includes: The feature vector is normalized and divided according to a ratio of 8:

2.

7. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1, characterized in that: The model building module uses a support vector machine as a classifier, and when training a translation error detection model according to the training set and the test set, it includes: Select the SVM kernel function and use the optimization strategy to find the optimal model parameters of the translation error detection model; Wherein, the SVM kernel function includes a linear kernel function and a radial basis kernel function; The optimization strategies include grid search optimization strategy and genetic algorithm optimization strategy; The model parameters include a penalty factor C and a kernel function parameter gamma.

8. The SVM-based machine-assisted English-Chinese technical translation error analysis system according to claim 1, characterized in that: The error detection module inputs the machine translation text to be analyzed into the trained translation error detection model, and after obtaining the error result of the machine translation text, it also includes: An error analysis report is generated according to the error result, wherein the error analysis report includes the original text of the sentence where the error is located, a machine translation version, an error type description, an error cause analysis, and modification suggestions.

9. A method for analyzing errors in machine-assisted English-Chinese technical translation based on SVM, applied to a system for analyzing errors in machine-assisted English-Chinese technical translation based on SVM as claimed in any one of claims 1 to 8, characterized in that: include: Constructing a professional technical text corpus, the professional technical text corpus includes sentence-aligned English original texts and Chinese translations, and using a machine translation engine to translate the English original texts to obtain Chinese machine translation texts; Establishing a marking rule, and marking errors of the Chinese machine translation text based on the marking rule, wherein the error marking information includes the error type, error location and correct translation; Preprocessing the Chinese translation and the Chinese machine translation text, and extracting key features from the preprocessed Chinese translation and the Chinese machine translation text, wherein the key features include lexical features, grammatical features and semantic features; Converting the key features and the error labeling information into feature vectors, and dividing the feature vectors into a training set and a test set; Using a support vector machine as a classifier, training a translation error detection model based on the training set and the test set, and adjusting parameters of the translation error detection model using a cross-validation method; The machine translated text to be analyzed is input into the trained translation error detection model to obtain the error result of the machine translated text.

Images