A parser based on the Construction Grammar formalism [](#1)[](#2), used for deep semantic natural language processing. The system is intended to enable easy exploration of novel parsing strategies and grammar engineering expressivity.
The current `cooking` grammar is developed to parse underspecified instructions from the kitchen domain into interpretations of the ontological model [SOMA-SAY](https://ease-crc.github.io/soma/). These consist of series of scenes and state transitions specifying the schemas evoked by the input sentence.
Inspired by the Embodied Construction Grammar framework [](#3), scg is intended to allow both the constructions' meaning poles as well as the analysis process to employ ontological knowledge. This facilitates evoking unspecified parameters (whose values might need to be inferred by later processing steps), helps disambiguating unclear instructions, and might also enable more informed search processes.
Scg's underlying mechanism is based on the unification and merging algorithms implemented in FCG [](#4), while employing simplified internal data structures and semantic output in the form of semantic triples, i.e. [unit feature value] pairs, with the aim to facilitate a tight integration into knowledge base-focused understanding pipelines.
<aid="1"></a>Fillmore, C. (1988). The Mechanisms of “Construction Grammar”. Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society, pp. 35-55. <aid="2"></a>Goldberg, A. E. (1995). Constructions: A Construction Grammar Approach to Argument Structure. Cognitive Theory of Language and Culture Series. <aid="3"></a>Feldman, J., Dodge E., & Bryant J. (2009). A Neural Theory of Language and Embodied Construction Grammar. The Oxford Handbook of Linguistic Analysis, B. Heine and H. Narrog, eds., pp. 111-138. <aid="4"></a>Steels, Luc, De Beule, Joachim. (2006). Unify and Merge in Fluid Construction Grammar. Symbol Grounding and Beyond, pp. 197-223.