With the rapid development of science and technology, new technical means are constantly being introduced in the field of education to improve the quality and efficiency of teaching. In the English teaching evaluation, deep learning, as an advanced machine learning method, has shown strong potential. Deep learning models can learn and extract features from a large number of complex data by simulating the working mechanism of human brain neural networks, thus achieving an accurate evaluation of learners' abilities. In English teaching, these models can analyze students' oral expression, writing level, reading comprehension and other skills, and provide objective and comprehensive evaluation. In recent years, due to the popularity of the Internet, a large number of English learning data have been collected, such as online homework, oral language recording, online testing, etc., which provides rich training materials for deep learning models. By learning these data, these models can capture the subtle differences of learners in language use, thus providing more accurate evaluation results and helping teachers to develop personalized teaching plans. MATLAB simulation shows that under the condition of certain evaluation criteria, the accuracy of oral ability assessment and the efficiency of oral ability assessment of higher vocational English by the deep learning scoring model are better than those of traditional English teaching.