Next: Machine Readable Dictionaries
Up: Computational Issues
Previous: Lexical Disambiguation
Any attempt to represent and disambiguate word senses depends heavily
on the representation of lexical semantic information. Many of the
early NLP systems relied on hand-coding of the lexicon, but this was
quickly realised to be problematic for the development of large-scale
systems. Research turned to development of automated techniques for
encoding of the lexicon. The initial attempts in this area were made
by basing lexica on electronic versions of dictionaries, Machine
Readable Dictionaries (MRDs). However, the need for frequency and
co-occurrence information as argued for by mcroy:92 and
copestake_briscoe:95 points to the need to augment lexica with
information which can be derived only through corpus analysis. In
this section, I will review several attempts to extract lexica from each
of these sources.