Are there available datasets quantifying semantic similarity for word pairs in English?
Many thanks for your help!
Dear Massimo Stella
Please take a look at the following links, these might be helpful:
Article Experimental data for computing semantic similarity between ...
Conference Paper SemEval-2012 task 6: a pilot on semantic textual similarity
Conference Paper SemEval-2015 Task 2: Semantic Textual Similarity, English, S...
Kind regards,
Cecilia-Irene Loeza-Mejía
Dear Massimo Stella , yes, there are several. I think to date, the most famous for semantic similarity in general is still the STS Benchmark dataset:
http://ixa2.si.ehu.eus/stswiki/index.php/STSbenchmark
https://paperswithcode.com/sota/semantic-textual-similarity-on-sts-benchmark
And on the word level those are particularly relevant:
http://www.gabormelli.com/RKB/WordSim-353_Dataset
http://www.gabormelli.com/RKB/Rubenstein-Goodenough_(RG-65)_Dataset
http://www.gabormelli.com/RKB/Miller-Charles_(MC-30)_Dataset
http://www.gabormelli.com/RKB/SimLex-999_Dataset
Hi Massimo! This one has similarities / relatedness for a large number of word pairs: https://staff.fnwi.uva.nl/e.bruni/MEN
I would also soon be able to share with you the individual-level similarity ratings of this: Preprint Structural differences in the semantic networks of younger a...
Fantastic, thank you all for your kind responses and for providing such great resources! I look forward to reading all these papers.
You can find a fairly comprehensive list here (p.9-10): https://arxiv.org/pdf/1801.09536.pdf
Simlex-999 (only 999 word pairs)
https://fh295.github.io/simlex.html
word2vec embeddings (can compute similarity for pretty much all words in the lexicon)
http://meshugga.ugent.be/snaut//
You can also use this space (from a news and wiki English corpus) to compute similarities: http://psicoee.uned.es/quantumlikespace/especifications/ASSE_CompareWordToWord.aspx
You just have to make a GET request with the word pairs as:
http://psicoee.uned.es/quantumlikespace/Service.svc/webHttp/CompareWordToWord?code=code&a=WORD1&b=WORD2
Regards
https://psyarxiv.com/6fuhv/
Hundreds of thousands of word pairs in running narratives
Kind of late to answer this, but just in case, I think these two sources might be helpful:
1) https://smallworldofwords.org/en/project/research which is described in the following paper:
De Deyne, S., Navarro, D.J., Perfors, A. et al. The “Small World of Words” English word association norms for over 12,000 cue words. Behav Res 51, 987–1006 (2019). https://doi.org/10.3758/s13428-018-1115-7
2) The USF Free association norms: http://w3.usf.edu/FreeAssociation/AppendixB/index.html described in this paper:
Nelson, D.L., McEvoy, C.L. & Schreiber, T.A. The University of South Florida free association, rhyme, and word fragment norms. Behavior Research Methods, Instruments, & Computers 36, 402–407 (2004). https://doi.org/10.3758/BF03195588
To anybody who still finds this valuable, I attach two links for databases providing concept property norms (1) and semantic similarity estimates derived from various corpuses using LSA, word2vec, and BERT algorithms (2)
1) https://cslb.psychol.cam.ac.uk/propnorms
2) http://lsa.colorado.edu/
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts