The New Oxford Dictionary of English (NODE) is out, to the delight of some and dismay of others. It is based on study of corpora, revealing language as it is used rather than as it "should" be. For instance, it accepts "data" as a singular mass noun and "they" as a singular pronoun, and takes no stand against split infinitives. The Daily Telegraph describes it as "a dumbed-down version of the OED." [Michael B Quinion , alt.usage.english, 26Aug98.]

Copyediting-l is a list for copy editors and other defenders of the English language. Send a "subscribe copyediting-l your name" message to , NEW-LIST, 03Aug98.]

Jamie Henderson says he has gotten "a surprisingly positive response" to his COLING-ACL'98 computational linguistics paper on syntactic parsing using a connectionist architecture with temporal synchrony variable binding (in the manner of SHRUTI). This generalizes in a structural way (as opposed to most connectionist nets using simple vector-based representations, without nested structures or variable binding). The paper -- Henderson and Lane 1998, at -- evaluates this learning method on a real corpus of naturally occurring text, with results that approach state-of-the-art statistical methods. Henderson says that connectionist methods can solve real problems involving ad-hoc independence assumptions, sparse data, etc.; these methods can be justified on purely empirical/engineering grounds; and connectionist NLP researchers will be taken seriously if they justify their work this way. [, connectionists, 14Aug98.]