Hacker News Reader

In case anybody's interested, one of our (former) graduate students has done some really fascinating work on computational analysis of prenominal modifiers (adjectives, etc.). For example, her ACL paper from 2011:

Mitchell, M., Dunlop, A., and Roark, B. (2011). Semi-Supervised Modeling for Prenominal Modifier Ordering. Proceedings of ACL 2011. http://www.aclweb.org/anthology/P/P11/P11-2041.pdf

View on HN

Interesting conclusion from the paper:

[...] for ordering prenominal modifiers, [...] a simple n-gram model outperforms position-specific models

This seems to imply that adjective order can be better described with common phrases rather than a set of definite grammar rules.

Thus the challenge of modern syntactic theory - to most it is clear there is some structure occurring, but attempts to quantify it are still beaten by rather naive statistical approaches.

Reply on news.ycombinator.com