HEADY: News headline abstraction through event pattern clustering

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 1243–1253, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics HEADY: News headline abstraction through event pattern clustering Enrique Alfonseca Google Inc. ealfonseca@google.com Daniele Pighin Google Inc. biondo@google.com Guillermo Garrido∗ NLP & IR Group at UNED ggarrido@lsi.uned.es Abstract This paper presents HEADY: a novel, ab- stractive approach for headline generation from news collections. From a web-scale corpus of English news, we mine syntac- tic patterns that a Noisy-OR model gener- alizes into event descriptions. At inference time, we query the model with the patterns observed in an unseen news collection, identify the event that better captures the gist of the collection and retrieve the most appropriate pattern to generate a head- line. HEADY improves over a state-of-the- art open-domain title abstraction method, bridging half of the gap that separates it from extractive methods using human- generated titles in manual evaluations, and performs comparably to human-generated headlines as evaluated with ROUGE. 1 Introduction Motivation. News events are rarely reported only in one way, from a single point of view. Dif- ferent news agencies will interpret the event in dif- ferent ways; various countries or locations may highlight different aspects of it depending on how they are affected; and opinions and in-depth anal- yses will be written after the fact. The variety of contents and styles is both an op- portunity and a challenge. On the positive side, we have the same events described in different ways; this redundancy is useful for summarization, as the information content reported by the majority of news sources most likely represents the central part of the event. On the other hand, variability and subjectivity can be difficult to isolate. For some applications it is important to understand, given a collection of related news articles and re- ∗Work done during an internship at Google Zurich. • Carmelo and La La Party It Up with Kim and Ciara • La La Vazquez and Carmelo Anthony: Wedding Day Bliss • Carmelo Anthony, actress LaLa Vazquez wed in NYC • Stylist to the Stars • LaLa, Carmelo Set Off Celebrity Wedding Weekend • Ciara rocks a sexy Versace Spring 2010 mini to LaLa Vasquez and Carmelo Anthony’s wedding (photos) • Lala Vasquez on her wedding dress, cake, reality tv show and fianc´e, Carmelo Anthony (video) • VAZQUEZ MARRIES SPORTS STAR AN- THONY • Lebron Returns To NYC For Carmelo’s Wedding • Carmelo Anthony’s stylist dishes on the wedding • Paul pitching another Big Three with “Melo in NYC” • Carmelo Anthony and La La Vazquez Get Married at Star-Studded Wedding Ceremony Table 1: Headlines observed for a news collection reporting the same wedding event. ports, how to formulate in an objective way what has happened. As a motivating example, Table 1 shows the dif- ferent headlines observed in news reporting the wedding between basketball player Carmelo An- thony and actress LaLa Vazquez. As can be seen, there is a wide variety of ways to report the …

Content truncated to 500 words