Brown, Simon (2016) Two measures of linguistic distance. Cultural Anthropology and Ethnosemiotics, 2 (3). pp. 2-12.

PDF (Published Version) - Published Version Restricted to Repository staff only

Abstract

Linguistic relatedness is often assessed on the basis of lexical analysis using the normalised Levenshtein distance (LDN), but a variant of this is the LDN divided (LDND) has been used in a similar way. As LDND is the LDN normalised by the ‘global distance’ (Γ(α, β)) between two languages, it is useful to consider both the properties of both LDND and Γ(α, β). Because Γ(α, β), like LDN, can not be greater than 1, LDND is ‘almost always’ greater than LDN and has no upper limit. However, for Indo-European word lists LDND is linearly related to LDN (p < 0.001) because Γ(α, β) has a very narrow distribution. Similar Γ(α, β) of were obtained in two numerical experiments based on (i) randomly generated ‘words’ or (ii) English words. This indicates that LDN may a better measure of lexical distance.

Item ID:	52553
Item Type:	Article (Research - C1)
ISSN:	2411-6459
Keywords:	distribution; global distance; Levenshtein distance; lexical distance
Date Deposited:	22 Mar 2018 04:50
FoR Codes:	47 LANGUAGE, COMMUNICATION AND CULTURE > 4704 Linguistics > 470403 Computational linguistics @ 100%
SEO Codes:	97 EXPANDING KNOWLEDGE > 970120 Expanding Knowledge in Language, Communication and Culture @ 100%
Downloads:	Total: 3
	More Statistics