PhenomenaTop
Contents
This page is intended to collect together various endeavours in the DELPH-IN community relating to documenting and exploring linguistic phenomena within DELPH-IN grammars. Please add your project to the list or add any missing information to any of the existing entries.
This project is an attempt at working towards an ‘encyclopedia’ of semantic analyses developed in the ERG. For more information see the ErgSemantics page.
This project is aiming to create a corpus annotated with linguistic phenomena independently of existing grammars for the purposes of investigating techniques into automatically detecting linguistic phenomena within grammars. An English phenomenon corpus based on DeepBank and focusing on syntactic and morphosyntactic phenomena is currently being developed by NedLetcher, TimBaldwin and EmilyBender. For further details and background, please see the paper Constructing a Phenomenal Corpus: Towards Detecting Linguistic Phenomena in Precision Grammars in the proceedings of the Workshop on High-level Methodologies for Grammar Engineering 2013.
A dedicated page will be coming soon. Contact NedLetcher for more information.
Typediff is a tool for exploring the types employed in the processing of input to DELPH-IN grammars. Typediff works by extracting the type identifiers found in the final AVM of a successful parse and sorting this bag of types according to HPSG-theoretic type categories such as 'sign' and 'synsem'. This data (which is not otherwise readily available) can be further refined through two different means in order to identify types correlated with linguistic phenomena of interest. One is the ability to perform a 'diff' of the set of types extracted from two different sentences: a sentence exemplifying a phenomenon and an otherwise similar sentence not exhibiting this phenomenon. The other feature is that types can be sorted in ascending order of coverage over a treebank, with the assumption being that more phenomenon-specific types will occur less frequently in a corpus.
The idea behind this project was for grammar writers to choose 100 phenomena covered by their grammar and provide illustrative examples in IGT for each phenomenon. The intention is to facilitate cross-grammar comparison, grammar exploration and potentially even better grammar documentation with phenomenon labels on constraints. Originally discussed at the Suquamish summit, currently there is only a partial catalogue for the Wambaya grammar. Grammar maintainers interested in developing a catalogue are very much encouraged to do so.
Home | Forum | Discussions | Events