On Finite-State Tonology with Autosegmental Representations

Building finite-state transducers from written autosegmental grammars of tonal languages involves compiling the rules into a notation provided by the finitestate tools. This work tests a simple, human readable approach to compile and debug autosegmental rules using a simple string encoding for autosegmental representations. The proposal is based on brackets that mark the edges of the tone autosegments. The bracket encoding of the autosegments is compact and directly human readable. The paper also presents a usual finite-state transducer for transforming a concatenated string of lexemes where each lexeme (such as ”babaa|HH”) consists of a segmental substring and a tonal substring into a chronological master string (”b[a]b[aa]”) where the tone autosegments are associated with their segmental spans.