Query-Focused Opinion Summarization for User-Generated Content

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 1660–1669, Dublin, Ireland, August 23-29 2014. Query-Focused Opinion Summarization for User-Generated Content Lu Wang1 Hema Raghavan2 Claire Cardie1 Vittorio Castelli3 1Department of Computer Science, Cornell University, Ithaca, NY 14853, USA {luwang, cardie}@cs.cornell.edu 2LinkedIn, CA, USA hraghavan@linkedin.com 3IBM T. J. Watson Research Center, Yorktown Heights, NY 10598, USA vittorio@us.ibm.com Abstract We present a submodular function-based framework for query-focused opinion summarization. Within our framework, relevance ordering produced by a statistical ranker, and information coverage with respect to topic distribution and diverse viewpoints are both encoded as submodular functions. Dispersion functions are utilized to minimize the redundancy. We are the first to evaluate different metrics of text similarity for submodularity-based summarization methods. By experimenting on community QA and blog summariza- tion, we show that our system outperforms state-of-the-art approaches in both automatic evaluation and human evaluation. A human evaluation task is conducted on Amazon Mechanical Turk with scale, and shows that our systems are able to generate summaries of high overall quality and information diversity. 1 Introduction Social media forums, such as social networks, blogs, newsgroups, and community question answering (QA), offer avenues for people to express their opinions as well collect other people’s thoughts on topics as diverse as health, politics and software (Liu et al., 2008). However, digesting the large amount of information in long threads on newsgroups, or even knowing which threads to pay attention to, can be overwhelming. A text-based summary that highlights the diversity of opinions on a given topic can lighten this information overload. In this work, we design a submodular function-based framework for opinion summarization on community question answering and blog data. Question: What is the long term effect of piracy on the music and film industry? Best Answer: Rising costs for movies and music. ... If they sell less, they need to raise the price to make up for what they lost. The other thing will be music and movies with less quality. ... Other Answers: Ans1: Its bad... really bad. (Just watch this movie and you will find out ... Piracy causes rappers to appear on your computer). Ans2: By removing the profitability of music & film companies, piracy takes away their motivation to produce new music & movies. If they can’t protect their copyrights, they can’t continue to do business. ... Ans4: It is forcing them to rework their business model, which is a good thing. In short, I don’t think the music industry in particular will ever enjoy the huge profits of the 90’s. ... Ans6: Please-People in those businesses make millions of dollars as it is!! I don’t think piracy hurts them at all!!! Figure 1: Example discussion on Yahoo! Answers. Besides the best answer, other answers also contain relevant information (in italics). For example, the sentence in blue has a contrasting viewpoint compared to the other answers. Opinion summarization has previously been applied to restricted domains, such as product reviews (Hu and Liu, 2004; Lerman et …

Content truncated to 500 words