When Did that Happen? — Linking Events and Relations to Timestamps

Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, pages 185–193, Avignon, France, April 23 - 27 2012. c⃝2012 Association for Computational Linguistics When Did that Happen? — Linking Events and Relations to Timestamps Dirk Hovy*, James Fan, Alfio Gliozzo, Siddharth Patwardhan and Chris Welty IBM T. J. Watson Research Center 19 Skyline Drive Hawthorne, NY 10532 dirkh@isi.edu, {fanj,gliozzo,siddharth,welty}@us.ibm.com Abstract We present work on linking events and flu- ents (i.e., relations that hold for certain periods of time) to temporal information in text, which is an important enabler for many applications such as timelines and reasoning. Previous research has mainly focused on temporal links for events, and we extend that work to include fluents as well, presenting a common methodol- ogy for linking both events and relations to timestamps within the same sentence. Our approach combines tree kernels with classical feature-based learning to exploit context and achieves competitive F1-scores on event-time linking, and comparable F1- scores for fluents. Our best systems achieve F1-scores of 0.76 on events and 0.72 on flu- ents. 1 Introduction It is a long-standing goal of NLP to process natu- ral language content in such a way that machines can effectively reason over the entities, relations, and events discussed within that content. The ap- plications of such technology are numerous, in- cluding intelligence gathering, business analytics, healthcare, education, etc. Indeed, the promise of machine reading is actively driving research in this area (Etzioni et al., 2007; Barker et al., 2007; Clark and Harrison, 2010; Strassel et al., 2010). Temporal information is a crucial aspect of this task. For a machine to successfully understand natural language text, it must be able to associate time points and temporal durations with relations and events it discovers in text. ∗The first author conducted this research during an in- ternship at IBM Research. In this paper we present methods to estab- lish links between events (e.g. “bombing” or “election”) or fluents (e.g. “spouseOf” or “em- ployedBy”) and temporal expressions (e.g. “last Tuesday” and “November 2008”). While previ- ous research has mainly focused on temporal links for events only, we deal with both events and flu- ents with the same method. For example, consider the sentence below Before his death in October, Steve Jobs led Apple for 15 years. For a machine reading system processing this sentence, we would expect it to link the fluent CEO of(Steve Jobs, Apple) to time duration “15 years”. Similarly we expect it to link the event “death” to the time expression “October”. We do not take a strong “ontological” position on what events and fluents are, as part of our task these distinctions are made a priori. In other words, events and fluents are input to our tempo- ral linking framework. In the remainder of this pa- per, we also do not make a strong distinction be- tween relations in general and fluents in particu- lar, and use them interchangeably, since our focus is only on the specific types of …

Content truncated to 500 words