Processing Temporal Information in Unstructured Documents

para a obtenção do grau de doutor em Informática (especialidade Ciência da Com-putação). 2012 Temporal information processing has received substantial attention in the last few years, due to the appearance of evaluation challenges focused on the extraction of temporal information from texts written in natural language. This research area belongs to the broader field of information extraction, which aims to automatically find specific pieces of information in texts, producing structured representations of that information, which can then be easily used by other computer applications. It has the potential to be useful in several applications that deal with natural language, given that many languages, among which we find Portuguese, extensively refer to time. Despite that, temporal processing is still incipient for many language, Portuguese being one of them. The present dissertation has various goals. On one hand, it addresses this current gap, by developing and making available resources that sup-port the development of tools for this task, employing this language, and also by developing precisely this kind of tools. On the other hand, its purpose is also to report on important results of the research on this area of temporal processing. This work shows how temporal processing requires and benefits from modeling different kinds of knowledge: gram-matical knowledge, logical knowledge, knowledge about the world, etc. Additionally, both machine learning methods and rule-based approaches are explored and used in the development of hybrid systems that are capable of taking advantage of the strengths of each of these two types of approach