Some Thoughts on Interface Structure(s)

There has recently been a very interesting public exchange of views regarding the relationship between MT and linguistics (Kettunen 1986; Slocum 1986). Kettunen – responding to Slocum's (1985) article – criticises MT for the fact that it sacrifices its reliance on "linguistically sound principles" to the goal of producing acceptable translations cost-effectively. For this reason, "the results may not have much to say to computational linguistics or to linguistics proper." Slocum's reply is that to a great extent "linguistic theory fails to account for linguistic fact", that "MT system developers have less time to engage in theorizing than they might like", and that "theoretical linguists (...) seldom if ever engage in application development".