Computer Algorithm Reconstructs Ancient Languages Better Than Humans
Program opens up the possibility of large-scale recreation of certain dialects.
A new study reveals that a computer program by researchers in Canada and California may be able to recreate extinct languages almost as well as a human. The reconstruction of dead languages – known as proto or root languages – is a meticulous time-consuming task for any linguist so this program could be a labour-saving mechanism that may assist in language analysis.
Manual reconstruction by a person requires looking at a word in two or more languages and making educated guesses about what the possible root word may have sounded like. This program uses an algorithm to do a similar thing, taking theories and rules about linguistics to configure possible root words. Tested against words already determined by linguists, the computer had an 85% success rate of being within one character of the ancient word.
Lead author Alexandre Bouchard-Côté, a statistician at the University of British Columbia admits in the journal Nature that the program is still rudimentary by comparison to manual labor, but can be improved upon:
Our system uses only a subset of the factors taken into consideration by a linguist, so we feel most of the discrepancies reflect things to be improved in our method.
The possibility of large-scale automatic reconstruction would aid in studying how words have evolved from protolanguage to our modern one and could tell us how people spoke in the past. Given the increased interest in preserving languages, the speed at which the program works and its potential to help understand past languages and history could prove to be quite essential.