Sunday, December 8, 2013

A tool for cognate research

We're almost there!
It is a matter of weeks before we say goodbye to the old site and introduce the new one. Just few details left... One of the biggest challenges has been the "Tools" section. As mentioned earlier this year on another post, one of Palaeolexicon's many roles is to provide you with tools that will help your research. One of the most painstaking things in historical linguistics, is the identification of cognates across languages. You know how it is... Open every single dictionary, lookup for words that sound/look relatively similar, have a relatively close semantic context and satisfy all kind of phonological rules. Now look at this screenshot below.

This is the "Cognate research tool". It will use various algorithms to identify possible cognates across many languages. It will save you lots of time, since it will do most of the job (including a semantic analysis). Once those cognate candidates are presented, it is up to you to evaluate the results. If something looks interesting enough, you can request further analysis on that candidate to find more distant relationships.

However, you should be cautious. Those algorithms have not been thoroughly tested and cannot guarantee safe results. The last decision is always yours. By time the algorithms will be improving and the data available for analysis will increase. There is this idea of creating plugins that will test the results against the phonogical rules that apply for every language. That should increase the accuracy significantly, however this feature belongs to a future project.