Application of LSTM in Assisting Personalized Learning for English Word Classification

At present, the Internet has entered the era of big data, and a large amount of information is constantly generated, such as text, audio, pictures, videos and other information, among which text information is the largest in quantity and the most easily accessible to people. The way text information is generated has caused the phenomenon of text information clutter, which is difficult to distinguish and organize manually due to the huge amount of information. Therefore, it becomes quite important to classify and organize textual information effectively and quickly, which gives rise to text classification techniques. Text classification is a classical research direction in the field of natural language processing, and the traditional research includes pre-processing of text, extraction of text features, training of machine learning classifiers, etc. With the great progress of deep learning technology in image recognition, machine translation and other fields, deep learning models have been proved to have great advantages in data pre-processing and feature extraction. In this paper, by studying the forward dependence problem of LSTM models, an attention-based combined forward and inverse LSTM model is designed to train the feature vectors of forward English text and inverse English text, respectively, and then combine them into the final feature vectors to be trained with classifiers. The Bi-LSTM model is also introduced as a comparison model to explore the influence of contextual information on English text classification. The final experimental results prove that the approach of this paper, which integrates the influence of contextual information and input sequences on the output, can significantly improve the accuracy of English text classification.