Sentiment Classification

Mark Liberman has more on that profile of Deborah Cameron which prompted me to investigate just what Cameron meant by Darwinist. He applies the notion of sentiment classification, the idea that one can gauge how a writer feels about the subject by applying an algorithmic approach. Often, this involves counting up the number of characteristic words or short word sequences: in a book review, “a must” tends to indicate a positive overall impression, whereas “don’t waste” is indicative of a negative opinion. (A breakdown of Amazon reader reviews, comparing the comment text to the numerical ranking, indicates that you’d much rather see “grisham” in the review of your novel than “predictable.”) Indicator words vary among settings, and what is negative or positive in one context is not necessarily weighted the same in another.

Liberman applies this to Ed Caesar’s profile of Cameron, noting that Caesar uses snippets like “rather prosaically,” “pet peeves,” “irked” and “bludgeoning its brains out” — all of which tend to indicate that the writer of the article has a somewhat negative view of the subject.

Of course, the same technique is applicable to whatever Cameron says, too. In fact, after working through some background material, I’d say that Darwinist and Darwinian are, for her, negative indicators, much as they seem to be for anybody outside of British evolutionary biology departments.

One thought on “Sentiment Classification”

  1. That may wel be true, but while I don’t consider myself a ‘Darwinist’ in the American sense of an implied belief system (the ‘church of Darwin’, ‘true believers in Darwin’, etc.) I do claim, with some pride, that my thought is Darwinian. The difference matters, and it would be a shame if the latter term came into sufficient disrepute that I’d have to disassociate myself from it as well. Charles Darwin isn’t a god, and he got some things wrong, but he got a hell of a lot more things right.

Comments are closed.