TimeML Temporal Expressions Detection for Catalan Using Semantic Roles and Semantic Networks

Nowadays, representation and processing of temporal information in natural language is receiving a great research interest. The main annotation scheme for representing temporal information is TimeML, which has been adopted as standard by a large number of researchers. However, available TimeML resources are very limited, specially in languages other than English. In this work we analyze the usage of semantic networks and semantic roles from a multilingual perspective for the automatic detection of temporal expressions following TimeML specifications. This approach has been evaluated for Catalan obtaining an 83.7% strict F-beta= 1, and compared to its results for English and Spanish confirming that it can be successfully applied in different languages.