Automated Comparison of Natural Languages - Software and Datasets of the Dissertation (Thesis)
Ansis Ataols Bērziņš · Zenodo (CERN European Organization for Nuclear Research) · 2019
In the dissertation few methods of definition and calculation of distance between natural languages are offered. Those methods are statistical and theoretically could be used for any human (spoken) language pairs, however in the dissertation they are approbated on dialects of Baltic languages. The methods are provided both for speech recordings or phonetical transcriptions. The concept of (all) phoneme space is introduced and metrics in it are defined. Also, for the purpose of expert evaluation of the results, the hierarchical choice metrics, which describes a distance between binary trees created by hierarchical categorization, are introduced. In this package scripts developed or modified during the work on the dissertation as well as datasets used by these scripts (apart speech WAV files) are included. Anybody is free to modify and use all contents of this package for research or educational purposes unless included in a commercial product. The name and the contact information of the developer should not be removed in any circumstances and modifications.