James, Let's put it this way: I know as much about computational linguistics as I do about advanced math, if you saw my comment to Eric.
I don't believe language can be quantified the way that most mathematicians and computational folks would like. My approach to syntax has three parts: base rules, transformations, and what I call social convention. This last is a fairly nebulous category, but basically it means that there are certain aspects of language usage that we can only explainby appealing to the fact that a society chooses to do it that way. I don't really have any examples from Hebrew yet, but in biblical Greek a couple of examples would be the "historic present" that we see all over the place in Mark, and the way that neuter plural subjects take a singular verb. On 23 Jun 2010 at 22:21, James Christian wrote: > > Hi, so it seems we agree that even if we are to perfectly define a > constituent driven CFG it is > destined to over generate. I appreciate that you have a strict > policy of separating syntax and > semantics but from an engineering perspective the approach has > little importance when > compared with the results. This is not to say I have anything > against implementing a generative > grammar with a strict separation between syntax and semantics it is > just that it is still not clear to > me how exactly this kind of over generation will eventually be > eliminated. > I don't know if you have looked into knowledge based expert systems. > This seems to be the kind > of approach you are suggesting. They are a very grey area in the > computational linguistics > literature. That is to say it is possible to find papers outlining > experiments with high or near > perfect performance in some very limited domain but wide coverage > remains elusive. You may or > may not have heard of the cyc project. A project which aims to > encode a basic knowledge of the > universe, something like what we call common sense. > In any case, with Hebrew I guess we'll just have to suck it and see. > For me, it is not the method > so much that carries any particular importance. It is the > performance of the system. That is to > say, what it all boils down to is how the system performs in terms > of over and under generation. > James Christian > > On 22 June 2010 18:58, <[email protected]> wrote: > > > On 21 Jun 2010 at 8:20, James Christian wrote: > > > > > Ok. Any ideas how you would implement this such that the > grammar > > would not over generate? > > Have you made any implementations of your scheme? James > Christian > > As I said, the constraints you are talking about are in the semantic > component. We can > take a phrase like Chris Rice's famous song "Smell the Color Nine" > and see that it's > syntactically well-formed, but semantic nonsense. My dad used to > jokingly use the phrase > "Is you speaking to I?" This one is a little more problematic, but > in a sense is still > syntactically well-formed at a deep grammar level. It includes a > subject, verb and > preposition with object. But there are agreement and case problems, > obviously. As I see it, > those are surface level aspects, i.e. the clause becomes badly > formed by filling the slots > with deliberate mismatches: "You" in English takes a different form > of the verb, and "I" in > English is used exclusively as a nominative. Where such distinctions > exist, they may or > may not be part of the actual syntactic component of grammar; I > haven't really thought that > through yet. > > Dave Washburn > > http://www.nyx.net/~dwashbur > _______________________________________________ > b-hebrew mailing list > [email protected] > http://lists.ibiblio.org/mailman/listinfo/b-hebrew > > > Dave Washburn http://www.nyx.net/~dwashbur _______________________________________________ b-hebrew mailing list [email protected] http://lists.ibiblio.org/mailman/listinfo/b-hebrew
