Application of English translation system design based on semantic relation classification in teaching ecological health

Abstract At present, the teaching ecology in the public environment is not healthy. The relationship between the ecological subjects (teachers and students) is indifferent and lacks resonance. Students' resistance to teaching leads to ineffective teaching. Meanwhile, the accuracy of English machine translation is generally not high, and there are problems such as insufficient accuracy of the semantic classification and unclear semantic expression. To improve the model robustness and solve the semantic classification problem in the NMT task, this paper proposes a lightweight bidirectional LSTM network SAT-BiLSTM (Bi-directional Long Short-Term Memory) with an attention mechanism. The SAT-BiLSTM translation network is composed of six decoders and six encoders, and the bidirectional LSTM network with an improved strong attention mechanism is used to process the input text and the output text. Experimental results show that compared with other NMT models, the system improves the translation speed and accuracy, achieves better statistical results on rare words translation, and the semantic classification is more accurate. This study is helpful to improve the effectiveness of English classroom teaching and create a healthy college English classroom ecosystem.