Similar-Associated-Both Test Collection (State of the art)

From ACL Wiki
Jump to navigation Jump to search
  • the contrast between taxonomical similarity (co-hyponymy) and association (co-occurrence)
  • 144 word pairs labeled similar-only, associated-only, or similar+associated
  • 48 pairs in each of the three classes
  • test collection created by Chiarello et al. (1990)
  • Chiarello et al. (1990) used the dataset in human priming experiments; they did not measure classification accuracy
  • dataset is provided in the Appendix of Chiarello et al. (1990); also available on request from Peter Turney
  • see also: Similarity (State of the art), SimLex-999 (State of the art)

Samples

Word pair Class label
table:bed similar
music:art similar
hair:fur similar
house:cabin similar
cradle:baby associated
mug:beer associated
camel:hump associated
cheese:mouse associated
ale:beer both
uncle:aunt both
pepper:salt both
frown:smile both

Table of results

Algorithm Reference Type Accuracy 95% confidence
Dual-Space Turney (2012) corpus-based 61.1% 52.6-69.1%
PairClass Turney (2008) corpus-based 77.1% 70.1-84.3%

Notes

References

Chiarello, C., Burgess, C., Richards, L., & Pollock, A. (1990). Semantic and associative priming in the cerebral hemispheres: Some words do, some words don't . . . sometimes, some places. Brain and Language, 38, 75{104.

Turney, P.D. (2008). A uniform approach to analogies, synonyms, antonyms, and associations. Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), Manchester, UK, pp. 905-912.

Turney, P.D. (2012). Domain and function: A dual-space model of semantic relations and compositions, Journal of Artificial Intelligence Research (JAIR), 44, 533-585.