| Users’ comments data contains a wealth of information, reflecting the inherent laws of the user’s behavior. Opinion mining is the automation technology of mining and analyzing the huge amount of data. Sentiment analysis is a challenging problem in natural language processing(NLP), text analysis and computational linguistics. In this paper, we study the aspect-level opinion mining of Chinese from two aspects of coarse-grained opinion mining and fine-grained opinion mining. We have made preliminary exploration on the field of opinion mining.The method of machine learning is used to study the coarse-grained opinion mining. The BNB model was proposed to carry out coarse grained sentiment analysis. After the pre-processing operation of corpus through the model, feature extraction is performed by the combination of unigram words and bigram words, Then we do the feature dimension reduction processing.We use Naive Bayesian classifier to analysis the coarse grained sentiment.We have made experimental comparison of different classification algorithms and feature dimensions, BNB model achieves a better effect which the accuracy is 94%.In the research of fine grained opinion mining, keyword matching method is used to extract the feature attributes, the accuracy rate is higher than the method of dependency grammar. In this paper, we construct the Chinese emotion dictionary, summarize the existing dictionary and network dictionary, and construct the field emotion dictionary.Using the constructed emotion dictionary to make the fine grained opinion mining of Chinese comments. The emotion is divided into positive, slightly positive, neutral, slightly negative, negative. The experimental indicate that the result of classification have showed its efficiency which added domain dictionary.Coarse grained and fine grained teaching evaluation corpus was annotated artificially. Teaching evaluation system use this corpus and our opinion mining results and methods, design and implement a teaching evaluation system for opinion mining. The system use the teaching materials and from the two aspects of coarse grain size and fine grain size to opinion mining, then show the results which can provide support for teachers and schools to make decisions. |
