A Scoring System of Oral English-Chinese Translation Based on Frame Semantic Analysis

In recent years, there has been an increasing interest in speech evaluation. However, most of the automatic evaluation systems focus on evaluating the quality of retelling questions and reading questions. In our best knowledge, research evaluating the quality of English-Chinese oral translation questions have not been found. This paper proposed an automatic method for scoring English-Chinese oral translation. This method leveraged three indicators, including key words, general idea of sentences and fluency, as the basis of scoring. The key words indicator was evaluated by the synonym analysis at the lexical level to evaluate while the general idea of sentences indicator was evaluated by the framework semantic analysis at the sentence level. Then, fluency indicator was scored by the speed of the speech. The experimental results demonstrated that using Mel-frequency cepstral coefficients (MFCC) and the Hidden Markov Model (HMM) algorithm to identify the key words, the system achieved average recognition rate in answering keywords of 97.3%. There is a good possibility that the proposed method may be effective for key word recognition.