OntoNotes: A Unified Relational Semantic Representation

Abstract

The OntoNotes project is creating a corpus of large-scale, accurate, and integrated annotation of multiple levels of the shallow semantic structure in text. Such rich, integrated annotation covering many levels will allow for richer, cross-level models enabling significantly better automatic semantic analysis. At the same time, it demands a robust, efficient, scalable mechanism for storing and accessing these complex inter-dependent annotations. We describe a relational database representation that captures both the inter-and intra-layer dependencies and provide details of an object-oriented API for efficient, multi-tiered access to this data.

An annotated corpus for the analysis of VP ellipsis

“…These include parse trees (Marcus et al 1993), thematic roles, word senses and ontological links (Pradhan et al 2007), co-reference relations (Weischedel and Brunstein 2005), discourse relation annotations based on Rhetorical Structure Theory (Carlson et al 2001), discourse graphs in GraphBank (Wolf and Gibson 2005), and connective-based discourse annotation for the entire corpus in the Penn Discourse Treebank (Prasad et al 2008). A thorough investigation of VPE in the Wall Street Journal texts is a much needed complement to these existing resources, and an ideal initial annotation project because of the possibility of utilizing existing resources together with the annotation in future work.…”

Any questions? Automatic question detection in meetings

“…And just as the ASR errors can degrade classification performance, so, too, can the parsing ones. Table VIII shows classification performance results using parse trees on the Broadcast Conversation subset of release 2.9 of the OntoNotes data set (LDC2009E05) [18]. In addition to reference transcriptions (derived from closed captions), the data contains manually annotated parse trees of some 13,000 utterances.…”

An Open Linguistic Infrastructure for Annotated Corpora

“…A fundamental tenet of the LAF model is that all annotations are in stand-off format, with references to primary data or other annotations. 24 . The Graph Annotation Format (GrAF) [13,14], the XML serialization of the model, is intended to function in much the same way as an interlingua in machine translation, that is, as a "pivot representation into and out of which user-and tool-specific formats are transduced, so that a transduction of any specific format into and out of GrAF accomplishes the transduction between it and any number of other GrAF-conformant formats.…”

“…The corpus closest to ANC-OLI in terms of richness of annotation and currency of language is the one million word English OntoNotes corpus [24], which includes annotations for Penn Treebank syntax, sense annotations using an in-house sense inventory, PropBank predicate argument structures, coreference, and named entities represented in a "normal form". As in MASC, all annotations have been hand-validated.…”