CU-TMP: Temporal Relation Classification Using Syntactic and Semantic Features

Proceedings of the 4th International Workshop on Semantic Evaluations (SemEval-2007), pages 129–132, Prague, June 2007. c⃝2007 Association for Computational Linguistics CU-TMP: Temporal Relation Classification Using Syntactic and Semantic Features Steven Bethard and James H. Martin Department of Computer Science University of Colorado at Boulder 430 UCB, Boulder, CO 80309, USA {bethard,martin}@colorado.edu Abstract We approached the temporal relation identi- fication tasks of TempEval 2007 as pair-wise classification tasks. We introduced a va- riety of syntactically and semantically mo- tivated features, including temporal-logic- based features derived from running our Task B system on the Task A and C data. We trained support vector machine models and achieved the second highest accuracies on the tasks: 61% on Task A, 75% on Task B and 54% on Task C. 1 Introduction In recent years, the temporal structure of text has be- come a popular area of natural language processing research. Consider a sentence like: (1) The top commander of a Cambodian resistance force said Thursday he has sent a team to recover the remains of a British mine removal expert kidnapped and presumed killed by Khmer Rouge guerrillas almost two years ago. English speakers immediately recognize that kid- napping came first, then sending, and finally saying, even though before and after never appeared in the text. How can machines learn to do the same? The 2007 TempEval competition tries to address this question by establishing a common corpus on which research systems can compete to find tempo- ral relations (Verhagen et al., 2007). TempEval con- siders the following types of event-time temporal re- lations: Task A Events1and times within the same sentence Task B Events1 and document times Task C Matrix verb events in adjacent sentences In each of these tasks, systems attempt to annotate pairs with one of the following relations: BEFORE, BEFORE-OR-OVERLAP, OVERLAP, OVERLAP-OF- AFTER, AFTER or VAGUE. Competing systems are instructed to find all temporal relations of these types in a corpus of newswire documents. We approach these tasks as pair-wise classifi- cation problems, where each event/time pair is assigned one of the TempEval relation classes (BEFORE, AFTER, etc.). Event/time pairs are en- coded using syntactically and semantically moti- vated features, and then used to train support vector machine (SVM) classifiers. The remainder of this paper is structured as fol- lows. Section 2 describes the features used to char- acterize event/time relations. Section 3 explains how we used these features to train SVM models for each task. Section 4 discusses the performance of our models on the TempEval data, and Section 5 sum- marizes the lessons learned and future directions. 2 Features We used a variety of lexical, syntactic and semantic features to characterize the different types of tempo- ral relations. In each task, the events and times were characterized using the features: word The text of the event or time words 1TempEval only considers events that occurred at least 20 times in the TimeBank (Pustejovsky et al., 2003) corpus for these tasks 129 Shhhhhhhhhhhhh QQQQ ((((((((((((( (((((PPhhhhh IN For NPhhhh …

Content truncated to 500 words