Research On Keyword Extraction And Sentiment Analysis For Chinese Text

With the rapid development of various social platforms,Natural Language Processing technology has become an inseparable part of life.The research of Chinese keywords is an important technology in many fields,such as machine translation,information retrieval and public opinion monitoring,and has long been a hot spot in the field of Natural Language Processing.However,current Chinese keyword extraction algorithms mostly use ready-made word segmentation tools to obtain candidate words,which leads to the performance of the algorithm greatly depends on the accuracy of pre-segmentation.In addition,most of the existing Chinese keyword sentiment analysis algorithms are based on the premise that keyword tags are provided in advance,but these expensive keyword tags are often not available in actual application scenarios.Although the pipeline method can be adopted to solve this problem,a number of related English works in recent years have fully confirmed that a joint model has greater potential.Therefore,in view of the above mentioned problems,this paper has conducted research on Chinese keyword extraction and end-to-end keyword sentiment analysis.The research content is mainly divided into two parts.Chinese keyword extraction algorithm.Aiming at the error problem caused by word segmentation error in the traditional Chinese keyword extraction algorithm,inspired by English key-phrase extraction method,this paper uses BIESO tag to label the boundary of keywords,and designs a keyword extraction algorithm based on Bi LSTM-CRF sequence labeling model.Considering the problem of neglecting the word information in the character-based sequence labeling method,this paper proposes a new word-related attention mechanism.Based on the principle of attention mechanism,this method focuses on the introduction of more relevant dictionary information,effectively uses the rich word information contained in large open source dictionaries,and enhances the ability of the model to capture keywords.According to the above improvements,a number of related experiments have been set up to prove the effectiveness of the integration of dictionary information in improving the performance of the sequence labeling keyword extraction method.End-to-End Chinese keyword emotion analysis algorithm.Aiming at the problems of complex training and neglect of joint information among sub tasks in the widely used pipelined Chinese keyword extraction and sentiment analysis methods in actual scenarios,this paper designs an end-to-end Chinese keyword sentiment analysis algorithm based on multi-task learning.The basic idea is to put two related sub tasks together for joint learning from the perspective of multi-task learning,so as to improve the generalization ability of the model.The model is implemented based on the Encoder-Decoder structure,and introduces contextual semantic information and the dependency information between keywords in the learning process,which provides a guarantee for the enhanced model to accurately capture the emotional changes of different keywords in complex Chinese expressions.