Grant: Research Tools and Instruments - Category 1 (<$150,000)

The field of Natural Language Processing (NLP) has diversified into a great many different application areas and approaches, and in order to clock overall progress and allow some consensus on common tools, the field has established a rigorous evaluation culture through so-called shared task competitions designed for many state of the art research topics. However, over the years, these competitions have required required the analysis of larger and larger corpora. For example, NIST-related collections have systematically increased their size over the years: in 2001, TIPSTER was about 3 MB, in 2002 AQUAINT was about 3GB, in 2006 Blogs-06 was about 150GB. Today, the Blogs-8 collection which is the new standard collection for the analysis of blogs is about 2.25TB spread across 30 millions documents. Any research team working on large-scale document analysis must therefore be able to accommodate these huge collections and must move at the terabyte level to keep up-to-date with recent trends. The CLaC Lab has a strong tradition in participating in shared task competitions and have been successful in various tracks. However, our lab, whose most recent machines are already four years old, work well with gigabyte level document collections, however, it cannot accommodate data collection at the terabyte level, The present proposal for equipment is intended to update our laboratory equipment to keep up with the current state of the art in large-scale document analysis. In order for us to continue to work with standard collections and continue our participation to these benchmark competitions, it has become urgent to update our facilities in order to keep our work at the international level and attract good graduate students.