Syntax and Semantics

From Geb

In this section, Hofstadter is (via Winograd) taking an implicit dig at generative-linguist theories that posit separate mental modules for syntax and semantics. This is all well and good, and there is evidence for both the Hofstadter/Winograd position, and for the separate module theory (I'm a bit out of the cogsci/linguistics loop, and I don't know if there have been slam dunks for one or the other position in the past five years).

I'm finding it annoying, however, that Hofstadter presents Winograd's comments on SHRDLU as if they are final proof for a unified syntax/semantics in natural language ("It is extremely interesting that in natural language, syntax and semantics are so deeply intertwined. ... But here, Winograd is telling us that-- at least when the usual sense of 'syntax' and 'semantics' are taken-- they merge right into each other, in natural language." (p. 631)). This is just one AI program-- albeit an apparently fairly successful one-- and there's no way that a single program can prove such a huge result about natural language in the human mind. (Also, I think that one could make still make a case for separate modules that are used simultaneously-- as Winograd describes-- in language processing).

Does anyone know if post-SHRDLU AI work has continued to parse sentences using both syntax and semantics simultaneously as Winograd describes?

-Aaron Stark