A Graph-Based Opinion Mining Approach for Reducing Information Loss and Overload in Product Reviews Analysis

Information overload is a real challenge that product designers face whiles trying to glean insight from online product reviews. Opinion summaries are limited in the richness of insight and text summarizations pose the risk of information loss for aspect based analysis. Although much effort has been spent yearly to advance the research in search, opinion analysis and text summarization, the same cannot be said for the provision of practical models and tools to leverage these advancements. In this work, we proposed a graph-based method centred on the Labelled Property Graph, sentiment analysis and text summarization. It indexes all tokens and opinions and allows for an explorative approach to aspect sentiment analysis whiles providing targeted sentence extracts for text summarizations through opinion-aware search. We show the limitations of text summarization for designers and show how our model can avoid them with expressive pattern matching and property filtering without reprocessing aspect sentiments.