Computational Linguistics

Computational Linguistics

Volume 8 / Number 10 / October, 1965 | HERBERT RUBENSTEIN AND JOHN B. GOODENOUGH
The study by Herbert Rubenstein and John B. Goodenough investigates the relationship between the similarity of context and the similarity of meaning (synonymy) of words. The researchers tested the hypothesis that the proportion of words common to contexts containing word A and contexts containing word B is a function of the degree to which A and B are similar in meaning. They used six different definitions of context, including all words within a sentence, content words within a frequency range, content words closely associated with the theme word, and words judged to be most closely associated with the theme word. The study involved 65 pairs of words ranging from highly synonymous to semantically unrelated, and judgments were made by two groups of college undergraduates. The results showed a positive relationship between the degree of synonymy and the overlap in contexts, but the slope of this relationship was steepest for the context defined by grammatical association. The study concluded that a pair of words is highly synonymous if their contexts show a relatively great amount of overlap, but inference of degree of synonymy from lesser amounts of overlap is uncertain. The findings suggest that the relationship between context similarity and synonymy is a linguistic universal, but the specific effects of different context definitions and the shape of the overlap versus synonymy function may be properties of the language materials and procedures used in the study.The study by Herbert Rubenstein and John B. Goodenough investigates the relationship between the similarity of context and the similarity of meaning (synonymy) of words. The researchers tested the hypothesis that the proportion of words common to contexts containing word A and contexts containing word B is a function of the degree to which A and B are similar in meaning. They used six different definitions of context, including all words within a sentence, content words within a frequency range, content words closely associated with the theme word, and words judged to be most closely associated with the theme word. The study involved 65 pairs of words ranging from highly synonymous to semantically unrelated, and judgments were made by two groups of college undergraduates. The results showed a positive relationship between the degree of synonymy and the overlap in contexts, but the slope of this relationship was steepest for the context defined by grammatical association. The study concluded that a pair of words is highly synonymous if their contexts show a relatively great amount of overlap, but inference of degree of synonymy from lesser amounts of overlap is uncertain. The findings suggest that the relationship between context similarity and synonymy is a linguistic universal, but the specific effects of different context definitions and the shape of the overlap versus synonymy function may be properties of the language materials and procedures used in the study.
Reach us at info@study.space