Chris Powell's Online Research

home | research | papers | resources | Advanced Interfaces Group

To follow ...

  • This section is under construction - but look at my thesis abstract to get a feel for my interests.

    Briefly however, my research interests are in the field of Natural Language Processing (NLP), particularly in specifying and using representations of Sense within a Categorial Grammar framework to construct logical representations of text.

    Some research highlights to date include:

    • Identification of a set of noun (verb) senses within the WordNet hypernym/hyponym taxonomy, dubbed Specialisation Classes and based on a class's information content, which allow the set of noun (verb) classes to be hugely abstracted with minimal loss of sense distinction.

    • Demonstration that for incremental parsing, the standard CKY chart parser cannot always correctly parse a sentence where there is single-term/compound-term ambiguity in the sentence. An extended (actually, the general form of a) chart parser (the Parallel-Shift Parser) is proposed.

    • Demonstration that inclusion of sense information into a Categorical Grammar framework:
      • appears to reduce rather than increase the number of calculations necessary to parse a sentence.

      • appears to promote incremental parsing in this otherwise non-incremental system.

      • allows the structural, semantic and sense disambiguation of a sentence in a cognitive manner, that is, one that falls foul of the Garden Path Effect.