Transformer-based a Automatic Scoring Model for Translation Jobs
Yiqing Lin · 2023
In the field of natural language processing, evaluating translation quality has been a long-standing challenge. Traditionally, the assessment of translation quality has relied heavily on manual assessment, which can be time-consuming, subjective and resource-intensive. These limitations spur the need for automated methods to streamline processes and increase objectivity and efficiency. To solve this problem, this paper introduces an automatic scoring system based on the Transformer model for evaluating the quality of translation tasks. The Transformer is a powerful sequence-to-sequence model that has achieved significant success in the field of natural language processing. By using the Transformer model, we can transform the translation task into a machine learning problem and automatically analyze the differences between the target language and the reference language. The training process of this automatic scoring system involves using a large amount of parallel corpora for data training, enabling the model to learn effective translation rules and semantic representations. In the testing phase, the system can accept the translation text to be evaluated and assign it an automatic score, representing the quality of its translation. This automatic scoring system can improve the efficiency and accuracy of translation tasks while reducing reliance on manual evaluation.