Thanks to Daniel Lemire for point to this recent paper by Peter Turney on “A Uniform Approach to Analogies, Synonyms, Antonyms, and Associations“.
Daniel Lemire consider this paper an example of the “more data beats better algorithms” principle most famously espoused by Google Director of Research Peter Norvig.
My take is a bit different. One message I heard repeatedly at the recent NSF Symposium on Semantic Knowledge Discovery, Organization and Use is that semantic researchers need to reduce their problem space to make progress. Peter is doing exactly that in his own work by taking what are perceived as distinct problems and generalizing them in order to treat them uniformly. Perhaps the broader community could profit from his approach and, um, learn by analogy.
3 replies on “Learning by Analogy”
Reducing the problem space. Nice.
LikeLike
Thanks, Daniel. I have an ambitious vision of showing how “all meaning comes from analogies”, but my own paper is no more than a small step in that direction.
LikeLike
Well, it’s at least a step in the right direction. And it’s more plausible (if less ambitious) step than the approaches proposed at the symposium to reduce all semantic problems to entailment or paraphrase.
LikeLike