My challenge has been to describe phoneme alternations using a method that is as simple and easy to understand as possible while still retaining its general applicability. Phonemic alternations appear in many languages in the inflection of words. For instance in Finnish a t to d consonant change occurs between rata - ‘track’ (nominative singular) and radoilla - ‘tracks’ (adessive plural), as does a vowel change from a to o. Similarly the Finnish word miekka (‘sword’) is the equivalent of the Estonian word mõõk, or the standard Finnish word kolme (three) is the equivalent of kolome in Eastern Finnish dialects. Linguists know what kinds of sounds alternate with each other and under what conditions they occur, but finding a precise formulation for these regularities and applying them by computer is far from easy.

I am trying to find an easier, more objective way of identifying and formulating observed rules and regularities in comparison to traditional methods. It is possible because two-level rules are independent of each other and each can be separately found. This is simple for a person, and my aim is the linguistically valid automatic deduction of rules on the basis of example material chosen by the linguist. I will implement the rules with Helsinki Finite-State Transducer Technology (HFST), which was developed by FIN-CLARIN, at which point we will be able to easily use groups of rules in both directions and combine them with other groups of rules.