Timeline extraction using distant supervision and joint inference

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1936–1942, Austin, Texas, November 1-5, 2016. c⃝2016 Association for Computational Linguistics Timeline extraction using distant supervision and joint inference Savelie Cornegruta Department of Biomedical Engineering King’s College London, UK savelie.cornegruta@kcl.ac.uk Andreas Vlachos Department of Computer Science University of Sheffield, UK a.vlachos@sheffield.ac.uk Abstract In timeline extraction the goal is to order all the events in which a target entity is involved in a timeline. Due to the lack of explic- itly annotated data, previous work is primar- ily rule-based and uses pre-trained temporal linking systems. In this work, we propose a distantly supervised approach by heuristically aligning timelines with documents. The noisy training data created allows us to learn mod- els that anchor events to temporal expressions and entities; during testing, the predictions of these models are combined to produce the timeline. Furthermore, we show how to im- prove performance using joint inference. In experiments in the SemEval-2015 TimeLine task we show that our distantly supervised approach matches the state-of-the-art perfor- mance while joint inference further improves on it by 3.2 F-score points. 1 Introduction Temporal information extraction focuses on extract- ing relations and events along with the time when they were true or happened. In this work we focus on timeline extraction, following the recent SemEval TimeLine shared task (Minard et al., 2015). The aim of the task is to extract timelines from multiple doc- uments consisting of events in which a given target entity is the main participant. An example timeline for the entity Steve Jobs extracted from 4 documents is given in Fig.1. The development data provided by the TimeLine shared task does not contain annotations for the var- ious intermediate processing stages needed, only a set of documents with annotated event mentions (in- put) and the timelines extracted for a few target en- tities (output). No training data was provided, thus participating systems used rules combined with tem- poral linking systems trained on related tasks in or- der to anchor events to temporal expressions and en- tities to construct the timelines. We propose a new approach to timeline extraction that uses the development data provided as distant supervision to generate noisy training data (Craven and Kumlien, 1999; Mintz et al., 2009). More specifically, we heuristically align the target entity and the timestamps from the timelines with auto- matically recognized entities and temporal expres- sions in the documents. This noisy labeled data set allows us to learn models for the subtasks of an- choring events to temporal expressions and to en- tities, without requiring training models on addi- tional data. Also, we improve the performance us- ing joint inference for both anchoring subtasks. In our experiments, we show that our distantly super- vised approach matches the state-of-the-art perfor- mance while joint inference further improves on it by 3.2 F-score points. Our code is publicly available at http://github.com/savac/timeline. 2 Timeline extraction The task of timeline extraction given a target entity and a set of documents can …

Content truncated to 500 words