KUL: Recognition and Normalization of Temporal Expressions

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 325–328, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics KUL: Recognition and Normalization of Temporal Expressions Oleksandr Kolomiyets, Marie-Francine Moens Department of Computer Science Katholieke Universiteit Leuven {oleksandr.kolomiyets, sien.moens}@cs.kuleuven.be Abstract In this paper we describe a system for the recognition and normalization of temporal expressions (Task 13: TempEval-2, Task A). The recognition task is approached as a classification problem of sentence con- stituents and the normalization is imple- mented in a rule-based manner. One of the system features is extending positive an- notations in the corpus by semantically similar words automatically obtained from a large unannotated textual corpus. The best results obtained by the system are 0.85 and 0.84 for precision and recall re- spectively for recognition of temporal ex- pressions; the accuracy values of 0.91 and 0.55 were obtained for the feature values TYPE and VAL respectively. 1 Introduction Recognition of temporal expressions1 is a task of proper identification of phrases with temporal semantics in running text. After several evalua- tion campaigns targeted at temporal processing of text, such as MUC, ACE TERN and TempEv- al-1 (Verhagen et al., 2007), the recognition and normalization task has been again newly reintro- duced in TempEval-2 (Pustejovsky & Verhagen, 2009). The task is defined as follows: determine the extent of the time expressions; in addition, determine the value of the features TYPE for the type of the temporal expression and its temporal value VAL. In this paper we describe the KUL system that has participated in this task. 1 Temporal expressions are sometimes referenced as time expressions and timexes. Architecturally, the system employs a pipe- lined information processing chain and imple- ments a number of machine learning classifiers for extracting the necessary information for the temporal value estimation. The normalization step employs a number of hand-crafted vocabula- ries for tagging single elements of a temporal expression and a rule-based system for estimat- ing the temporal value. The performance of the system obtained the values of 0.85 and 0.84 for precision and recall respectively for the recogni- tion of temporal expressions. The accuracy for the type and value is 0.91 and 0.55 respectively. The remainder of the paper is organized as follows: Section 2 reports on the architecture of the system with single modules and describes theirs functions. Section 3 presents the results and error analysis; the conclusions are provided in Section 4. 2 System Architecture The system is implemented in Java and follows a pipelined method for information processing. Regarding the problems it solves, it can be split in two sub-systems: recognition and normaliza- tion. 2.1 Recognition of Temporal Expressions This sub-system is employed for finding tempor- al expressions in the text. It takes a sentence as input and looks for temporal expressions in it. Pre-processing: At this step the input text un- dergoes syntactic analysis. Sentence detection, tokenization, part-of-speech tagging and parsing are applied2. Candidate selection: Since only certain lexi- cal categories can be temporal expressions and …

Content truncated to 500 words