A significant obstacle to the development of intelligent natural language processing systems is the lack of rich knowledge bases containing representations of world knowledge. For experimental systems it is common practice to construct small knowledge bases by hand; however, this approach does not scale well to large systems. An alternative is to attempt to extract the desired information from existing knowledge sources intended for human consumption; however, attempts to construct broad-coverage knowledge bases using in-depth analysis have met with limited success. In this paper we present some work on an alternative approach that involves using shallow processing techniques to build a hybrid knowledge representation that stores information in a partially analysed form.

This paper uses materials from annotation studies of lexical cohesion (Beigman Klebanov and Shamir, 2005) and of definite reference (Poesio and Vieira, 1998; Vieira, 1998) to discuss the complementary nature of the two processes.

Number 764

Abstract: This paper presents a new framework, “text as scene”, which lays the foundations for the annotation of two coreferential links: discourse deixis and bridging relations. The incorporation of what we call textual and contextual scenes provides more flexible annotation guidelines, broad type categories being clearly differentiated. Such a framework that is capable of dealing with discourse deixis and bridging relations from a common perspective aims at improving the poor reliability scores obtained by previous annotation schemes, which fail to capture the vague references inherent in both these links. The guidelines presented here complete the annotation scheme designed to enrich the Spanish CESS-ECE corpus with coreference information, thus building the CESS-Ancora corpus.