Đang chuẩn bị liên kết để tải về tài liệu:
Báo cáo khoa học: "Efficient Processing of Underspecified Discourse Representations"

Underspecification-based algorithms for processing partially disambiguated discourse structure must cope with extremely high numbers of readings. Based on previous work on dominance graphs and weighted tree grammars, we provide the first possibility for computing an underspecified discourse description and a best discourse representation efficiently enough to process even the longest discourses in the RST Discourse Treebank. | Efficient Processing of Underspecified Discourse Representations Michaela Regneri regneri@coli.uni-sb.de Saarland University Markus Egg egg@let.rug.nl t University of Groningen Alexander Koller a.koller@ed.ac.uk University of Edinburgh Abstract Underspecification-based algorithms for processing partially disambiguated discourse structure must cope with extremely high numbers of readings. Based on previous work on dominance graphs and weighted tree grammars we provide the first possibility for computing an underspecified discourse description and a best discourse representation efficiently enough to process even the longest discourses in the RST Discourse Treebank. 1 Introduction Discourse processing has emerged as a highly relevant source of information for applications such as information extraction and automatic summarisation Taboada and Mann 2006 outline this and further applications . But discourse structures cannot always be described completely either due to genuine ambiguity Stede 2004 or to the limitations of a discourse parser. In either case only partial information on discourse structure is available. To handle such information underspecification formalisms can be used. Underspecification was originally introduced in computational semantics to model structural ambiguity without disjunctively enumerating the readings and later applied to discourse parsing Gardent and Webber 1998 Schilder 2002 . However while the existing algorithms for underspecification processing work well for semantic structures they were not designed for discourse structures which can be much larger. Indeed it has never been shown that underspecified discourse reprentations UDRs can be processed efficiently since the general-purpose implementations are too slow for that task. In this paper we present a new way to implement and process discourse underspecification in terms of regular tree grammars RTGs . RTGs are used as an underspecification formalism in semantics Koller et al. 2008 .