Relational Model of Conceptual Distance between Bangla Words

Abstract

Words in a language are related to each other. This relation is based on their conceptual properties. (This paper avoids using the term “semantic property”, generally used by the contemporary NLP workers for measuring distance between words, the reason being that we employ different orientations behind the measurement of relatedness). Essentially, this work considers the psycho-sociological facts in the experiments, where a number of native speakers of Bangla manually suggests distance measurement between any two words. This work presents a statistical approach with a psycho-analytical elaboration for measuring the conceptual distance between words in terms of Bangla language. To be precise it calculates co-relations of the assessments collected through a survey among different individuals. A conceptual distance is used to suggest the implicit pragmatic nature of the Bangla words and it also implies an elementary taxonomy for Bangla words. As a result, the conceptual distance between Bangla words in the semantic field can very usefully be quantified and thus can be a crucial factor for a computational application like Bangla word net. Incidentally we find that there is a very high correlation (r = 0.95) between two different sets of human judgments and at the same time an assuringly high correlation (r = 0.95 being the upper limit) is observed when the respondents duplicated the same task with the same pairs of words at different points of times. This is a pioneering study in Bangla.

Acknowledgement

The survey used in this study was conducted with the help of Baidehi Sengupta and enormous supports for some other issues came from Rimi Ghosh Dastidar. We are thankful to them. We are also grateful to all the people who participated in the survey. We are grateful to the referee for constructive suggestions.

Notes

¹ This is a piece from “Sonar Tari” of Rabindranath Tagore and is translated by William Radice, (freely available on the internet).

² We have used data from Samsad Bengali-English dictionary digitised by the University of Chicago. In a personal correspondence, James Nye, the Bibliographer for Southern Asia, and Director, South Asia Language and Area Centre, The University of Chicago, informed us that there are approximately 20,950 headwords in this dictionary. See http://dsal.uchicago.edu/dictionaries/biswas-bengali/.