New Sentiment Word Detection In Web Texts And Key Sentiment Sentence Extraction

In recent years, text sentiment analysis has become a significant area in natural language processing. It can analyze the view, idea and emotion of posters and be widely used in areas such as product evaluation and opinion monitoring. Currently, traditional type of news is spreading through the Internet, which allows citizens to express and exchange their ideas on Weibo and blogs freely. In this paper, some work related to new sentiment word detection and chapter sentiment analysis is done based on web texts, experiments are conducted on the corpus of Chinese Opinion Analysis Evaluation. The results show the validity of the proposed methods, the main work in this paper are:(1)Extract new sentiment words in Chinese Weibo. We take Chinese Weibo as statistical object and find that traditional methods based on co-occurrence such as mutual information is limited. Therefore, a method based on context called CNG(Context N-Gram) is proposed to detect and extract new sentiment words, which features N-Gram and takes SVM as the classifier. The experiments show that the proposed method do better than co-occurrence methods, and large number of Chinese sentiment words are nouns which is a big obstacle for traditional methods.(2)Key sentiment sentence extraction and prediction. The sentiment of long texts or chapters is largely based on a few key sentences and the extraction and prediction towards them determines the analysis result. In this paper, a novel deep learning framework to predict sentiment distributions based on Recursive Neural Network as well as Support Vector Machine is introduced. There are sentiment features including not only grammar information such as sentiment and negative words, but also statistical information like word vector in deep learning. Meanwhile, chapter features like sentence pattern and position are also involved. Experiments show the objectivity of news has an effect on the analysis result, while key sentiment sentence extraction provides new methodology for text sentiment analysis.