In this monograph, the authors address the challenge of summarizing opinions from web-based blogs and reviews. Various approaches are considered, with the goal of locating opinions in blog threads that favor or oppose a particular topic and then separately summarizing those opinions. The results support the approach taken and indicate directions for further improvements.

The first section introduces the foundations of sentiment analysis and opinion mining, which are incorporated in a web-based context to generate opinion summaries, as opposed to content summaries. The next two sections describe related work in the literature and identify two datasets selected for experimentation, a collection of blog threads and a collection of bank reviews. The rest of the paper discusses approaches to generating meaningful opinion summaries in the blog and review spheres of the web. The goal is to combine opinion mining with summarization techniques to elicit mass opinion and to characterize pro and con arguments. The authors show that extending approaches by adding opinion analysis and semantic aspects improves the results significantly. A key insight across the results is that it is best to perform opinion mining before any summarization efforts, to avoid the loss of opinion based on fact. The authors propose future work that includes enhanced evaluation methods such as linguistic quality and user satisfaction. The paper’s 47 references cover the field well.

The authors have produced a substantive yet very readable paper that accurately reflects this area.