Recognition Method of Important Words in Korean Text based on Reinforcement Learning

Recognition Method of Important Words in Korean Text based on Reinforcement Learning Feiyang Yang Department of Computer Science and Technology, Yanbian University. 917936177@qq.com Yahui Zhao ∗ Department of Computer Science and Technology, Yanbian University. 903873610@qq.com Rongyi Cui Department of Computer Science and Technology, Yanbian University. cuirongyi@ybu.edu.cn Abstract The manual labeling work for constructing the Korean corpus is too time-consuming and labo- rious. It is difficult for low-minority languages to integrate resources. As a result, the research progress of Korean language information processing is slow. From the perspective of represen- tation learning, reinforcement learning was combined with traditional deep learning methods. Based on the Korean text classification effect as a benchmark, and studied how to extract im- portant Korean words in sentences. A structured model Information Distilled of Korean (IDK) was proposed. The model recognizes the words in Korean sentences and retains important words and deletes non-important words. Thereby transforming the reconstruction of the sentence into a sequential decision problem. So you can introduce the Policy Gradient method in reinforce- ment learning to solve the conversion problem. The results show that the model can identify the important words in Korean instead of manual annotation for representation learning. Further- more, compared with traditional text classification methods, the model also improves the effect of Korean text classification. 1 Introduction The languages of ethnic minorities have created the diversity of Chinese characters and are an impor- tant part of Chinese characters, providing important support for the development of national culture. However, the research on Korean natural language processing in my country is still in the development stage, and the related research is still relatively lagging behind South Korea and North Korea(Bi, 2011). For manual annotation of Korean sentences, the structure division requires a lot of energy and time. So for this problem, we associate the method of representation learning. Representation learning has been widely used in text classification, sentiment analysis, language reasoning and other fields in recent years. It is a basic problem in the field of artificial intelligence, and it is particularly important in the field of natural language processing. Therefore, we use this method as the core logic of the model, aiming at Ko- rean text, identifying important words and performing sentence classification tasks on newly constructed sentences. The resulting structural representation does not require manual annotation, greatly reducing manpower and scientific research resources. In order to find important Korean words in sentences, we use the effect of text classification as feed- back in reinforcement learning. The current mainstream text classification models are roughly divided into four types: bag-of-words model, sequence model, structure representation model, and attention model. The bag-of-words representation model often ignores the order of words, such as deep average networks, self-encoders (Joulin A, 2017); the sequence representation model often only considers the words themselves, but ignores the phrase structure, such as CNN, RNN and other neural network models (Y, 2014); structural representation models often rely on pre-specified parse trees to construct structured representations, such as Tree-LSTM, recursive autoencoders (Zhu …

Content truncated to 500 words