On the unification of syntactic annotations under the stanford dependency scheme

Abstract

Several incompatible syntactic annotation schemes are currently used by parsers and corpora in biomedical information extraction. The recently introduced Stanford dependency scheme has been suggested to be a suitable unifying syntax formalism. In this paper, we present a step towards such unification by creating a conversion from the Link Grammar to the Stanford scheme. Further, we create a version of the BioInfer corpus with syntactic annotation in this scheme. We present an application-oriented evaluation of the transformation and assess the suitability of the scheme and our conversion to the unification of the syntactic annotations of BioInfer and the GENIA Treebank. We find that a highly reliable conversion is both feasible to create and practical, increasing the applicability of both the parser and the corpus to information extraction.

Parser Adaptation to the Biomedical Domain without Re-Training

“…Our models were tuned on a development set of 600 sentences and then evaluated on the remaining 500 sentence test set, using the same split as Pyysalo et al (2007a) and Rimmel and Clark (2009). The vocabulary in these sentences diverges considerably from that found in the WSJ, with about 27% of the tokens being unseen.…”

The Stanford typed dependencies representation

“…Recent work on parser evaluation using dependency graphs in the biomedical domain confirms that researchers regard dependency-based evaluation as a more useful surrogate for extrinsic task-based evaluation (Clegg and Shepherd, 2007;Pyysalo et al, 2007a). In their evaluation, Clegg and Shepherd (2007) aimed at analyzing the capabilities of syntactic parsers with respect to semantically important tasks crucial to biological information extraction systems.…”

“…In the biomedical domain, SD has recently been used in evaluations of parsers (Clegg and Shepherd, 2007;Pyysalo et al, 2007a). Pyysalo et al (2007a) assessed the suitability of the SD scheme over the Link Grammar dependency scheme in an application-oriented evaluation.…”

“…Pyysalo et al (2007a) assessed the suitability of the SD scheme over the Link Grammar dependency scheme in an application-oriented evaluation. The Link Parser indeed uses a very fine-grained set of relations, which often makes distinctions of a structural rather than a semantic nature.…”

“…The SD scheme, by aiming for an intermediate level of granularity, and targeting semantic dependencies, provides a more adequate representation for applications. Therefore, to increase the usability of the BioInfer corpus (Pyysalo et al, 2007b), which provides manually annotated data for information extraction in the biomedical domain and originally followed the Link Grammar scheme, Pyysalo et al (2007a) developed a version of the corpus annotated with the SD scheme. They also made available a program and conversion rules that they used to transform Link Grammar relations into SD graphs, which were then hand-corrected (Pyysalo et al, 2007b).…”