UC3M System: Determining the Extent, Type and Value of Time Expressions in TempEval-2

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 329–332, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics UC3M system: Determining the Extent, Type and Value of Time Expressions in TempEval-2 María Teresa Vicente-Díez, Julián Moreno Schneider, Paloma Martínez Department of Computer Science Universidad Carlos III de Madrid Avda. Universidad, 30 Leganés, 28911, Madrid, Spain. {tvicente, jmschnei, pmf}@inf.uc3m.es Abstract This paper describes the participation of Universidad Carlos III de Madrid in Task A of the TempEval-2 evaluation. The UC3M system was originally developed for the temporal expressions recognition and normalization (TERN task) in Spanish texts, according to the TIDES standard. Current version supposes an almost-total refactoring of the earliest system. Additionally, it has been adapted to the TimeML annotation schema and a considerable effort has been done with the aim of increasing its coverage. It takes a rule-based design both in the identification and the resolution phases. It adopts an inductive approach based on the empirical study of frequency of temporal expressions in Spanish corpora. Detecting the extent of the temporal expressions the system achieved Precision/Recall of 0.90/0.87 whereas, in a determining the TYPE and VALUE of those expressions, system results were 0.91 and 0.83, respectively. 1 Introduction The study of temporality in NLP is not a new task. However, in the last years it has witnessed a huge interest. Initiatives like TempEval task or the Automatic Context Extraction1 (ACE) TERN competitions have boosted research on the field and have promoted the development of new resources to the scientific community. There are two main advantages in participating in these evaluations. On the one 1 Automatic Content Extraction Evaluation. National Institute of Standards and Technology (NIST) http://www.itl.nist.gov/iad/mig//tests/ace/ hand it is possible to measure the systems’ performance under standardized metrics, sharing datasets and other resources. On the other hand, it is possible to make comparative evaluations among distinct participants looking forward the same objectives but using different approaches. Until recently, most of temporally annotated corpora, as well as temporal taggers, were available in English. Since languages as Spanish start to become prominent in the field it seems interesting the development of specific resources. Tempeval-2 has contributed to this target in a significant way thanks to the release of annotated corpora and the publication of specific guidelines (Sauri et al., 2009), (Saurí et al., 2010). This paper resumes the participation of the UC3M system in the task of determining the extent and resolving the value of time expressions in texts (Task A). This system was originally developed for the Spanish TERN task proposed in ACE 2007 evaluation (Vicente-Díez et al., 2007), achieving encouraging results although it was in a early stage of development. The system follows a ruled-based approach whose knowledge base has been inducted from the study of annotated temporal corpora (Vicente-Díez et al., 2008). A machine learning approach was initially discarded due to the limitation of annotated Spanish corpora. The aims of this work were to improve the coverage of the original system and test its …

Content truncated to 500 words