Workshop on Animacy and Information Status Annotation

Trustees Of Boston University, Boston

Investigators

Abstract

It is well known that in order to communicate effectively, it is not enough to speak in grammatical sentences. One needs to start from what the hearer knows or is interested in and get from there to the new information that one wants to convey. Understanding how human speakers do this is important as a pure research topic but it is also of great practical importance: enabling humans and machines to communicate in natural language will help with a variety of information transfer tasks, including tutoring, search, and translation. To make headway beyond the linguistic studies that are currently underway, a unified, universally accepted coding system for the phenomena that influence this part of language understanding is needed. In computational linguistics, uniform standards for marking text and annotating corpora are increasingly recognized as the key to progress; they allow researchers to make sure that they are consistent in the way they apply categories. They also allow the use of machine learning methods to help researchers to gain new insights and to analyze more texts more efficiently. This workshop project focuses on two aspects of this coding problem, an understanding of the notions of 'discourse status' (is the referent being mentioned new to the discourse or already familiar?) and of 'animacy' (e.g. is the referent a human, a machine, an insect, a building?). The workshop brings together linguists and computer scientists with the goal of creating a gold standard coding procedure for these two important categories of linguistic content. The findings will then be disseminated through the auspices of several large linguistic annotation projects that have ties to key computer science working groups.