Software finds the common roots of languages
It is expected to free linguists from the time-consuming calculations of evolutionary linguistics.
An algorithm developed by mathematicians in the United States and Canada seems to be able to trace the common roots of modern languages.
The researchers say that the software confirmed the linguists’ theories of the “proto-language” from which 637 modern languages originated in the region of Southeast Asia and the Pacific.
The algorithm, for example, found that the ancient word “bituqen” must have been the common root of the word “star” in all these languages, from the “kalokalo” of the Fiji Islands to the “mintol” of Taiwan and the “biten Of Borneo.
Push to study the genealogy of languages
Details of the algorithm, which is expected to significantly accelerate the study of the genealogy of languages, are published in the American review “PNAS”.
The researchers, led by statistician Alexander Bouchard-Kote of the University of Vancouver, used the rules that linguists have formulated about how word sounds evolve over the centuries.
The software analyzed a database of 120,000 words from various Pacific languages, along with their pronunciations and definitions. The algorithm then undertakes to design the evolutionary tree of all these languages, from their common root to the present day, calculating the probability that the sound of a word with a specific meaning has changed according to a specific rule.
To make sure that the program gives the correct, or at least quite possible, results, the researchers compared the resulting family tree with the corresponding phylogenetic tree created by linguists “by hand” after years of effort.
Identification in 85% of cases
Indeed, the results agreed in 85% of cases. In other cases, researchers suspect that the algorithm fell out because it took into account fewer rules than linguists. They assure, however, that there is considerable room for improvement.
If proven to be reliable, the software could free linguists from the time-consuming and tedious calculations of evolutionary linguistics.
Already, the program seems to confirm a theory of 1955, which until now was practically impossible to test. The analysis of the data for the 637 languages showed that, indeed, the sounds that are of particular importance in order to distinguish one word from another are more likely to remain unchanged over time.
Source: The Science Step
Posted: 12/02/2013 18:18
http://www.tovima.gr/science/technology-planet/article/?aid=497923&wordsinarticle=%CE%B1%CE%BB%CE%B3%CF%8C%CF%81%CE%B9%CE% B8% CE% BC% CE% BF