Graham Katz

Error Analysis of the TempEval Temporal Relation Identification Task

The task to classify a temporal relation between temporal entities has proven to be difficult with unsatisfactory results of previous research. In TempEval07 that was a first attempt to standardize the task, six teams competed with each other for three simple relationidentification tasks and their results were comparably poor. In this paper we provide an analysis of the TempEval07 competition results, identifying aspects of the tasks which presented the systems with particular challenges and those that were accomplished with relative ease.

Introduction to the Special Issue on Arabic Computational Linguistics

In recent years there has been a significant increase in the amount of research on the computational analysis of the Arabic language being carried out around the world. This includes both a steady growth of the computational linguistics research community in the Arab-speaking world, as evidenced by the emergence of scientific societies such as the Egyptian Society for Language Engineering1, as well as the existence of major initiatives on Arabic natural language processing (NLP) in Europe and North America (particularly in the United States by the DARPA-funded GALE program2). In addition, the past year alone saw the publication of two monographs on Arabic Computational Linguistics [Farghaly 2010; Habash 2010], a special issue of this journal [Shaalan and Farghaly 2009], a new release by the Linguistics Data Consortium of the Arabic Treebank [Maamouri et al. 2009] (among other Arabic-language resources), as well as a number of conferences and workshops on Arabic NLP around the world. At the 2010 Georgetown University Roundtable on Language3 (GURT), whose theme was Arabic Language and Linguistics, linguists from around the world and from a wide range of different linguistic traditions and sub-disciplines came together in Washington, D.C. to share their research on the Arabic language. A two-day special session dedicated to Arabic computational linguistics and a half-day preconference tutorial on Arabic NLP drew researchers from both academia and industry. The presentations in the special session spanned the range from applied NLP to natural language understanding, from automatic diacritization to speech act classification, providing attendees with a snapshot of the current state of Arabic NLP. The papers in this volume were selected from those presentations to represent the different approaches and techniques currently being used in the field of Arabic computational linguistics. Arabic NLP research faces a number of well-known challenges, due to both the morphological complexity of the Arabic language and to the particularities of Arabic orthography. These challenges have been addressed extensively in Arabic NLP research which has focused on Modern Standard Arabic (MSA) [Habash and Rambow 2005; Diab et al. 2007; Larkey et al. 2007; Buckwalter 2007; Dichy and Farghaly 2007]. Significant recent attention, however, has been paid to the problems for NLP