Research on Constructing Sentiment Dictionary of Online Course Reviews based on Multi-source Combination

Abstract

The construction of sentiment dictionary is an important task in text sentiment analysis. Using the sentiment words of specific fields to establish a domain-oriented sentiment dictionary can significantly improve the effect on sentiment recognition and classification in the specific field. A method of constructing a sentiment dictionary for online course reviews was proposed, which based on multi-source combination. In the first place, the sentiment words were identified and extracted using TextRank and the word2vec model, which combined with the general sentiment dictionary and the online course reviews corpus. Then, the label propagation algorithm was applied to discriminate the sentiment polarity of sentiment words, thus constructing a sentiment dictionary for online course reviews. Through the accuracy experiment on the determination of the sentiment polarity of words and the experiment on sentiment classification based on different dictionaries, the accuracy rate, the recall rate and the F value are calculated. The experimental results show that the proposed method was an accuracy and effective way to achieve the sentiment classification of online course reviews.