Title: Recent Developments in Sentiment Analysis: Categorization of Sentiment Views and Opinion Compounds

Abstract: In my talk, I will first address the notion of "sentiment views" by which the perspective of the holder of some opinion is meant. One commonly distinguishes between the view of the speaker of the utterance (i.e. speaker views) and the view of entities participating in an opinion event (i.e. actor views). I will discuss how opinion words can be effectively categorized according to their sentiment view. I discuss what features are required and show that this task can be achieved with few labeled training data.
In the second part of my talk, I will address "opinion compounds", which are noun compounds whose head is an opinion noun, such as "user rating" and "victim support". We focus on the categorization of the modifier which can represent an opinion holder or an opinion target.
This categorization is difficult since the context of noun compounds does not contain explicit cues as to the relation between head and modifier. We discuss which features are necessary in order to classify opinion compounds and also highlight the importance of sentiment views as an auxiliary classification.