Bernhard Wälchli

The consonant template in synchrony and diachrony

This paper investigates the importance of (unique) consonants in identifying cross-linguistic cognates. It is argued that consonants play a crucial role both in diachrony for the identification of cognates and in synchrony for the identification of stemforms. The same algorithm ― alignment of the consonant template (ACT) ― is applied both in diachrony for identifying cognates and in synchrony for aligning stemforms. It is argued that identifying cognates is essentially the alignment of their consonant templates. Since the alignment of consonant templates is over-generating, ACT must be strongly constrained by semantics. A method is presented to extract cognates directly from parallel texts which is exemplified and evaluated mainly on the basis of Lithuanian and Latvian. For identifying cognates, a three step procedure is applied: (a) finding semantically equivalent forms (SEF), (b) finding equivalent consonants (EC), and (c) alignment of the consonant template (ACT).

Keywords: Baltic languages, historical-comparative method, cognates, statistical natural language processing (NLP), consonants, autosegmental phonology, semantic equivalents, stemforms