[Prev][Next][Index]

Re: [xtag-meeting]: Question supertagging



Thanks,
I'll fwd to Pierre.
For some reason I thought that J. Chen used Srini's implementation ....

Libin said he's finishing his implementation now.  It seems interesting

Good luck in Baltimore !

> > Hi everyone,
> > Three quick questions from Pierre Boullier :
> >
> > -Apart from Srini's, are there any other supertaggers implemented ?
> > -If so, does anyone have a reference or know someone we could ask ?
>
> John Chen implemented one a couple of years ago. He has a paper in some conf,
> probably on his website as well. Recently, John, Srini and Mike Collins have
> used re-ranking to improve on an existing supertagger (either John's or a
> newly implemented one) -- the results were reported in TAG+ in Venice.
>
> More recently, Libin Shen (2nd year student of Joshi's) has new numbers on
> supertagging that beat previous results. His email is libin@linc.
>
> > -Has there been any attempt to do rule-based supertagging ?
>
> I think Srini tried to use Transformation Based Learning for supertagging. It
> was infeasible with the existing Brill implementation due to memory and time
> overruns as far as I can remember. I think he probably has a line in this
> thesis saying this.
>
> -A



References: