Search

This paper focuses on the contribution of temporal relations inference and distributional semantic models to the event ordering task. Our system automatically builds ordered timelines of events from different written texts in English by performing first temporal clustering and then semantic clustering. In order to determine temporal compatibility, an inference from the temporal relationships between events –automatically extracted from a Temporal Information Processing system– is applied. Regarding semantic compatibility between events, we analyze two different distributional semantic models: LDA Topic modeling and Word2Vec word embeddings. Both semantic models together with the temporal inference have been evaluated within the framework of SemEval 2015 Task 4 Track B. Experiments show that, using both models, the current State of the Art is improved, showing significant advance in the Cross-Document Event Ordering task. Este artículo se centra en estudiar la contribución que la inferencia de relaciones temporales y los modelos semánticos distribucionales hacen a la tarea de ordenación de eventos. Nuestro sistema construye automáticamente líneas de tiempo con eventos extraídos de diferentes documentos escritos en inglés. Para ello realiza primero una agrupación temporal y posteriormente una agrupación semántica. Para determinar la compatibilidad temporal se realiza una inferencia sobre las relaciones temporales entre los eventos extraídos de un sistema automático de procesamiento de información temporal. Para la compatibilidad semántica entre eventos hemos analizado dos modelos semánticos distribucionales distintos: LDA Topic Modeling y Word2Vec Word Embeddings. Ambos modelos semánticos junto con la inferencia temporal han sido evaluados bajo el marco de evaluación de SemEval 2015 Task 4 Track B. Los experimentos muestran que, usando ambos modelos se mejora el estado del arte actual, implicando un avance importante en la tarea de ordenación de eventos multidocumento. This paper has been partially supported by the Spanish government, project TIN2015-65100-R, project TIN2015-65136-C2-2-R and PROMETEOII/2014/001.

Automatic annotation of temporal expressions is a research challenge of great interest in the field of information extraction. Gold standard temporally-annotated resources are limited in size, which makes research using them difficult. Standards have also evolved over the past decade, so not all temporally annotated data is in the same format. We vastly increase available human-annotated temporal expression resources by converting older format resources to TimeML/TIMEX3. This task is difficult due to differing annotation methods. We present a robust conversion tool and a new, large temporal expression resource. Using this, we evaluate our conversion process by using it as training data for an existing TimeML annotation tool, achieving a 0.87 F1 measure -- better than any system in the TempEval-2 timex recognition exercise., Comment: Proc. LREC (2012)