Comparison of Dictionary Based and Corpora Based Phonological Inventories: Andic Data

Linguists distinguish between two types of frequencies: type and token frequency. Each of them is used for its own purpose and says something about its own aspect of the language. We are going to use these frequencies for quantitative phonological analysis of low-resource languages. This work focuses on the difference between automatically calculated sound frequencies in two types of data: corpora and dictionaries. Our hypothesis is that frequencies of sounds in them are correlated and the frequency calculated from the dictionary data can be used to describe phonology in the same way as the frequency calculated from the corpus data. To test it, we use corpora and dictionaries of Andic languages. Some of them already have corpora (Botlikh, Andi), some of them have texts in their grammatical descriptions from which we could make corpora (Bagvalal). For each language we will calculate frequencies of sound, and then compare them with each other to find a dependency. Ideally, they would be correlated. If this work can confirm our hypothesis, it may contribute to future quantitative studies of the phonology of low-resource languages since it will be possible to predict one value from another.