The score, as attentive readers of SC know, started when Mark Liberman attempted to neutralize the gender features of pronouns referring to this blog and its author.
Now, Andrew Sullivan has gone and made the argument that computational linguistics ought to produce better natural language understanding than regular reading. Ostensibly, the point was that in arguing that an essay by Josh Marshall didn't pay any attention to 9/11, he had not committed an egregious error by writing instead that Marshall didn't mention 9/11. In fact, some mention of the date itself had appeared, though not in the context intended by Sullivan. The "money quote", as Mr. Sullivan likes to say: "I wasn't engaged in linguistic computer analysis of the piece".
SC actually agrees with Andrew Sullivan's defense of himself, by appeal to linguistic pragmatics. While the truth conditions of Sullivan's original statement were not met -- Marshall did mention 9/11 -- the intent of Sullivan's claim was that it was not a subject of Marshall's essay, and that claim is valid (SC hesitates to call it true, on the grounds that arguing about the truth conditions of rhetoric is dicey). But Sullivan's specific statement with regards to computational linguistics is that only a good, solid parsing (maybe with the Linguist's Search Engine) could have uncovered the date string. SC therefore welcomes anybody who sent Andrew Sullivan an e-mail on this point to send him a link to the parsing tool they used. He suspects that, as usual with these challenges, no response is forthcoming, and also as usual, it won't only be due to the sample size. Chalk up another one for Careful Reading.
Comments