We present a system code-named OpSum for topic-based opinion summarization and sentiment analysis of mobile phone reviews. It enables users to decide whether to purchase or not based on a summary of the reviews for that mobile phone. Our system organizes the reviews based on product aspects extracted from the dataset and it also provides a sentiment analysis of these reviews. Selection of useful reviews is done from these collections of organized reviews. We discuss the effectiveness of using GRNNs and LSTMs for sentiment analysis and their trade-offs. We alsoperformed extractive summarization using Integer Linear Programming to extract the most distinct and important sentences from the clusters which will represent the reviews in the entire cluster.
展开▼
