Hi Ed,
Eduard Hovy schrieb:
SMT is re-treading the path of older approaches, but now doing things automatically that used to be done by hand:
- the initial IBM work recreated word-replacement MT, but learned the replacement rules automatically - Och's and other current SMT is redoing EBMT, but learning the phrases (i.e., examples) automatically - Yamada and Knight, Wu, and Melamed each are working on versions of transfer, with the rules, again, being learned
Well, one could also establish a similar list for EBMT. Generally, I would even claim that things can only be done automatically if one has done the task
before by hand, many, many times. So I am not surprised that SMT (and EBMT) lag behind rule-based approaches (at least conceptually).
wrt. hand-build or automatically acquired translation knowledge in EBMT:
also within the SMT literature there are papers on language models and other papers on translation models and it would be unfair to accuse someone (or the
community) not to tackle both in every paper. In the same way, in EBMT, there are papers on how to 'stitch' chunks together and other papers on how to
obtain the chunks automatically. By the way: in our book (Carl and Way, 2003) the majority of the technical chapters tackle the latter point!
why should EBMT be able to solve problems that even rule-based methods (i.e. 'by hand') cannot solve? You see the misconception:
The open questions of MT are still open: no-one can properly handle interpersonal/stylistic/pragmatic effects of communication. If one could use EBMT methods to capture style, for example, you'd be doing something that syntax-based transfer approaches would find very hard to do, and SMT approaches would struggle with given their need for large corpora of unitary style.
while SMT is automatizing MT approaches which are outdated in the rule-based paradigm already since 20 or 30 years, EBMT is expected to solve problems
which "no-one can properly handle".
But there is also an appropriate challenge in Ed's comment: EBMT implements an analogy-like top-down approach (see discussion last year on this list)
and nothing - in principle - prevents to take "interpersonal/stylistic/pragmatic effects of communication" as a guiding principle to find analogies in translated
texts. But, I guess, first one would have to do this 'by hand' and see exactly how it could work and then wait 20 years to automatise the task ...
Michael
_______________________________________________ MT-List mailing list [EMAIL PROTECTED] http://www.computing.dcu.ie/mailman/listinfo/mt-list
