This item is available under a Creative Commons License for non-commercial use only
Sentiment classification concerns the application of automatic methods for predicting the orientation of sentiment present on text documents. It is an important subject in opinion mining research, with applications on a number of areas including recommender and advertising systems, customer intelligence and information retrieval. SentiWordNet is a lexical resource of sentiment information for terms in the English language designed to assist in opinion mining tasks, where each term is associated with numerical scores for positive and negative sentiment information. A resource that makes term level sentiment information readily available could be of use in building more effective sentiment classification methods. This research presents the results of an experiment that applied the SentiWordNet lexical resource to the problem of automatic sentiment classification of film reviews. First, a data set of relevant features extracted from text documents using SentiWordNet was designed and implemented. The resulting feature set is then used as input for training a support vector machine classifier for predicting the sentiment orientation of the underlying film review. Several scenarios exploring variations on the parameters that generate the data set, outlier removal and feature selection were executed. The results obtained are compared to other methods documented in the literature. It was found that they are in line with other experiments that propose similar approaches and use the same data set of film reviews, indicating SentiWordNet could become an important resource for the task of sentiment classification. Considerations on future improvements are also presented based on a detailed analysis of classification results.
Ohana, Bruno, "Opinion mining with the SentWordNet lexical resource" (2009). Dissertations. Paper 25.