Semantic Dependency Parsing (SDP) is defined as the task of recovering sentence-internal predicate–argument relationships for all content words (Oepen et al., 2014; 2015; 2016).

SDP target representations, thus, are bilexical semantic dependency graphs. Two SDP shared tasks have been run as part of the 2014 and 2015 International Workshops on Semantic Evaluation (SemEval). In mid-2016, all task data, system results, and associated tools have been released publicly through the Linguistic Data Consortium (LDC). Additionally, a sub-set of the data is available under a more permissive license for direct download; please see the Access page (in the top or bottom menu) for further information.

The original SemEval tasks comprised three distinct target representations (over the same text), dubbed (a) DELPH-IN MRS-Derived Semantic Dependencies (DM), (b) Enju Predicate–Argument Structures (PAS), and (c) Prague Semantic Dependencies (PSD). The LDC release adds to this collection one additional type of bi-lexical semantic dependencies, viz. (d) Combinatory Categorial Grammar word–word dependencies (CCD), as well as additional background material (e.g. the English Resource Semantics and tectogrammatical trees, from which DM and PSD, respectively, were derived).

This site provides an archive of the original 2014 and 2015 SemEval task pages, combined with an interactive search interface and pointers to more recent releases of SDP data and general contact information.


Following are two example graphs (for a simplification of WSJ sentence #20209013), first in the DM representation. Semantically, technique arguably is dependent on the determiner (the quantificational locus), the modifier similar, and the predicate apply. Conversely, the predicative copula, infinitival to, and the vacuous preposition marking the deep object of apply can be argued to not have a semantic contribution of their own (for DM, there also is an on-line parsing service).

For the same sentence, the PSD graph has many of the same dependency edges (albeit using a different labeling scheme and inverse directionality in a few cases), but it analyzes the predicative copula as semantically contentful and does not treat almost as scoping over the entire graph.

XHTML 1.0 | Last updated: 2017-01-28 (15:01)