Word similarity dataset

Word similarity dataset is a dataset with similarities between words, often pairs of words. Such datasets are often used to evaluation tools that provide methods for lexical semantics, such as semantic similarity. Researchers have established several word similarity datasets.

Name	Year	Pairs	Range	Reference
Rubenstein-Goodenough (RG)	1965		0.02–3.94	Contextual correlates of synonymy
Miller-Charles (MG)	1991			Contextual correlates of semantic similarity
WordSim-353	2001	353		Placing search in context: the concept revisited
MEN	2012			Multimodal distributional semantics
SimLex-999	2014	999	0.23–9.80	SimLex-999: Evaluating Semantic Models with (Genuine) Similarity Estimation
SimVerb-3500	2016	3500		SimVerb-3500: A Large-Scale Evaluation Set of Verb Similarity

Quiz

For the RG read Contextual correlates of synonymy.

What value has the synonymy judgement between the word pair "coast" and "forest"?

The SimLex-999 may be read in Python with

from pandas import read_csv
import requests
from six import StringIO
from zipfile import ZipFile

url = "https://www.cl.cam.ac.uk/~fh295/SimLex-999.zip"
filename = 'SimLex-999/SimLex-999.txt'
simlex = read_csv(ZipFile(StringIO(requests.get(url).content)).open(filename), sep='\t')

What is the mean similarity rating?

External links

• http://wordvectors.org/
• https://aclweb.org/aclwiki/Similarity_(State_of_the_art)