Dear 何俊昊:

> The question is 'How do we make sure that we never get Syntax error
> (e.g. really robust glue rules)'. Here are my thoughts.
>
>    The most common syntax errors are shift/reduce conflict and
> reduce/reduce conflict.
>    I have a brief look at the document of GNU bison. I think the method
> mentioned in it is worthy of trying. When a conflict occurs, the
> parser will split into different parsers, one for each possible shift
> or reduction. So the parser can proceed as usual.
That's generalized LR parsing I think. Two problems: can easily get 
unmanageable (exponential with length) and there is no way to decide on 
a parse. Unless, of course, you use this forking during training, 
evaluate translations in some way, and then collect statistics on each 
conflict, to code the best decision for runtime.
>    And I have another thought. That is we can adopt a statistical
> model, just like Hidden Markov model adopted in the POS tagger. We can
> train the parser with some corpus and choose the most appropriate
> action to do in the parser based on the result of training. But this
> thought is not so specific so far.
In any case you would need to do something like the above: get all parse 
trees, translate for each parse tree, evaluate the translation obtained 
with some model, and then transfer these probabilities to rule 
probabilities in your parser. But then you would have to get all parses 
at runtime and score them again. This looks like research, not GSoC to 
me. And it departs from Apertium spirit (rule-based). So I prefer the above.

[snip]

Mikel


-- 
  Mikel L. Forcada (http://www.dlsi.ua.es/~mlf/)
Departament de Llenguatges i Sistemes Informàtics
Universitat d'Alacant
E-03071 Alacant, Spain
Phone: +34 96 590 9776
Fax: +34 96 590 9326


------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to