[developers] FSC Input to cheap

[David Henry Witting]

To provide some context for my prior message, my name is David Witting
and I am currently a student and soon-to-be graduate of the
Computational Linguistics program at the University of Washington, and
am working for Automata, Inc. as part of my path to graduation at the
end of this summer.  I'm using feature structure charts for my unknown
word handling right now, but I think they have a lot of potential for
other applications, but I'm still working on figuring out how to use
this tool properly.  For instance, in some cases I would like the
parser to treat a word as if it has a particular part of speech, or
utilizes a particular lexical entry, or perhaps even uses a generic
lexical entry even though the word exists in the lexicon.

Thanks for your help and I apologize for the double message,

David Witting

> I've been exploring the capabilities of the PET parser and have been
> making use of the Feature Structure Chart format. However, I'm having
> difficulty figuring out how the features in the FSC relate to the
> features in the grammar, primarily when it comes to formatting the FSC
> itself, and also (in some cases) getting the parser to adhere to the
> constraints that I define.

> My first and most substantial issue is the same as described here
> (http://lists.delph-in.net/archives/developers/2011/001573.html). That
> is, even though I have given a part of speech tag to a word, the
> parser is not constrained to use that part of speech, and will utilize
> any and all parts of speech that it finds in the lexicon.  Is there
> any way to override this behavior?

> Thank you,

> David Witting