A distant supervised learning system for the TAC-KBP Slot Filling and Temporal Slot Filling Tasks

This paper describes the system implemented by the NLP GROUP AT UNED for our first participation in the Knowledge Base Population at the Text Analysis Conference (TACKBP). For this Slot Filling Task, our approach was to design a distant supervised learning system, which was then specialized for the Regular Slot Filling and Full Temporal Slot Filling subtasks. From the initial Knowledge Base and source document collection distributed by the organizers, we automatically gathered training data for supervised slot classifiers. We used a rich document representation, augmenting syntactic dependency trees with named entity recognition, coreference resolution, temporal events annotation and semantic graph transformations.