Sentiment Analysis On The Rotten Tomatoes Movie Review Dataset

The Rotten Tomatoes movie review corpus is a collection of movie reviews collected by Pang and Lee in [2]. This corpus has been analysed in [3] where each sentence is parsed into its tree structure and each node is assigned a fine-grained sentiment label ranging from 1 − 5 where the numbers represent very negative, negative, neutral, positive and very positive respectively. In this paper we use this data on ath000 phrases and all the methods in this paper are assessed by training on a random subset of phrases (and their subphrases) of size approximately 4/5 of the data set and testing using the remaining 1/5. The idea is to use the non-associative functions and the parser trees structures to modify the feature vectors.