Michael Wiegand

Topic-Related Polarity Classification of Blog Sentences

Though polarity classification has been extensively explored at various text levels and domains, there has been only comparatively little work looking into topic-related polarity classification. This paper takes a detailed look at how sentences expressing a polar attitude towards a given topic can be retrieved from a blog collection. A cascade of independent text classifiers on top of a sentence-retrieval engine is a solution with limited effectiveness. We show that more sophisticated processing is necessary. In this context, we not only investigate the impact of a more precise detection and disambiguation of polar expressions beyond simple text classification but also inspect the usefulness of a joint analysis of topic terms and polar expressions. In particular, we examine whether any syntactic information is beneficial in this classification task.

Optimizing language models for polarity classification

This paper investigates the usage of various types of language models on polarity text classification - a subtask in opinion mining which deals with distinguishing between positive and negative opinions in natural language. We focus on the intrinsic benefit of different types of language models. This means that we try to find the optimal settings of a language model by examining different types of normalization, their interaction with smoothing and the benefit of class-based modeling.