Thanks, John. It's little surprise to me that allowing for these
reduced-relative constructions complicates parsing, for machines and
humans alike.
I am, though, interested in seeing what research has been done in the
NLP field with respect to these reduced-relative passives, such as how
accurately parsers or other systems recognize them and correctly
determine their voice.
Regards,
Sean
John F. Sowa wrote:
> Sean Igo wrote
>
> > I'm interested in an English garden-path phenomenon which
> > arises when verbs in passive voice occur in reduced relative
> > clauses, e.g.:
> >
> > "The horse _raced_ past the barn fell."
> >
> > as opposed to the non-reduced relative clause version,
> >
> > "The horse that _was raced_ past the barn fell."
>
> I recall some psycholinguistic studies from many years ago
> that indicate faster human response times when relative
> pronouns are used in such constructions (and in some related
> constructions in which they are optional). Unfortunately,
> I do not have the references handy. [...]
>
> For many broad-coverage parsers, the difference in time can be
> enormous. In fact, most of the speed up results from faster
> parsing of the sentences that do *not* contain any relative
> pronouns -- i.e., the time spent in testing for such options
> is longer than the time spent on the ones that actually occur.
This archive was generated by hypermail 2b29 : Tue Jan 02 2007 - 06:37:43 MET