A Sequence Labelling Approach to Quote Attribution

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 790–799, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics A Sequence Labelling Approach to Quote Attribution Tim O’Keefe† Silvia Pareti⋄ James R. Curran† Irena Koprinska† Matthew Honnibal‡ † e -lab, School of IT ⋄School of Informatics ‡Centre for Language Technology University of Sydney University of Edinburgh Macquarie University NSW 2006, Australia United Kingdom NSW 2109, Australia {tokeefe,james,irena}@it.usyd.edu.au S.Pareti@sms.ed.ac.uk matthew.honnibal@mq.edu.au Abstract Quote extraction and attribution is the task of automatically extracting quotes from text and attributing each quote to its correct speaker. The present state-of-the-art system uses gold standard information from previous decisions in its features, which, when removed, results in a large drop in performance. We treat the problem as a sequence labelling task, which allows us to incorporate sequence features without using gold standard information. We present results on two new corpora and an aug- mented version of a third, achieving a new state-of-the-art for systems using only realis- tic features. 1 Introduction News stories are often driven by the quotes made by politicians, sports stars, musicians, and celebri- ties. When these stories exit the news cycle, the quotes they contain are often forgotten by both read- ers and journalists. A system that automatically ex- tracts quotes and attributes those quotes to the cor- rect speaker would enable readers and journalists to place news in the context of all comments made by a person on a given topic. Though quote attribution may appear to be a straightforward task, the simple rule-based ap- proaches proposed thus far have produced disap- pointing results. Going beyond these to machine learning approaches presents several problems that make quote attribution surprisingly difficult. The main challenge is that while a large portion of quotes can be attributed to a speaker based on simple rules, the remainder have few or no contextual clues as to who the correct speaker is. Additionally, many quote sequences, such as dialogues, rely on the reader understanding that there is an alternating se- quence of speakers, which creates dependencies be- tween attribution decisions made by a classifier. Elson and McKeown (2010) is the only study that directly uses machine learning in quote attribution, treating the task as a classification task, where each quote is attributed independently of other quotes. To handle conversations and similar constructs they use gold standard information about speakers of previ- ous quotes as features for their model. This is an unrealistic assumption, since gold standard informa- tion is not available in practice. The primary contribution of this paper is that we reformulate quote attribution as a sequence labelling task. This allows us to use sequence features with- out having to use the unrealistic gold standard fea- tures that were used in Elson and McKeown (2010). We experiment with three sequence decoding mod- els including greedy, Viterbi and a linear chain Con- ditional Random Field (CRF). Furthermore we present results on two new cor- pora and …

Content truncated to 500 words