Next: Disclaimer
Up: The EP4IR grammar
Previous: Nested frames
We shall describe only a few of the many possible examples, leaving it to
the reader to contrive others.
- Information Retrieval
The grammar was developed for IR applications, in which the traditional
keywords are to be replaced by frames obtained from phrases. All frames
are first unnested, so that each document is represented by a bag of
frames without nesting. The frames are also morphologically normalized,
using a lemmatizer and the typing information provided. Furthermore,
semantically related frames will be clustered together (see [Koster et al., 1999]).
- Information Analysis and Modelling
An important step in most analysis techniques is to start out from an
informal (i.e. verbal) description of the problem domain or the problem,
and scan it for nouns and verbs: the nouns are candidate objects or
classes and the verbs are candidate methods [Abbott, 1983]. Other
elements, like the adjectives can also be used [Graham, 1994]. The HM frame
representation obtained by using EP4IR can be used straight away for this
purpose. The same technique can be used for the validation of an
existing Object-Oriented Analysis model [Frederiks et al, 1996].
- Question answering system
A question answering system of any kind will need more information
than is included in the HM frames, e.g. quantifiers and determiners.
Of course the grammar does already recognize these constructs, so by
modifying three or four lines in the transduction they will also be
expressed. Luckily, the grammar already knows how to parse questions.
Feel free to use the AGFL system for your purposes according to the
GPL/LGPL license. Let us know if you have a nice application.
For more complicated projects you might consider collaborating with
the authors of this paper.
Next: Disclaimer
Up: The EP4IR grammar
Previous: Nested frames
Kees Koster
2002-05-01