A Probabilistic Approach to Diachronic Phonology

Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 887–896, Prague, June 2007. c⃝2007 Association for Computational Linguistics A Probabilistic Approach to Diachronic Phonology Alexandre Bouchard-Cˆot´e∗ Percy Liang∗ Thomas L. Griffiths† Dan Klein∗ ∗Computer Science Division †Department of Psychology University of California at Berkeley Berkeley, CA 94720 Abstract We present a probabilistic model of di- achronic phonology in which individual word forms undergo stochastic edits along the branches of a phylogenetic tree. Our ap- proach allows us to achieve three goals with a single unified model: (1) reconstruction of both ancient and modern word forms, (2) discovery of general phonological changes, and (3) selection among different phyloge- nies. We learn our model using a Monte Carlo EM algorithm and present quantitative results validating the model. 1 Introduction Modeling how languages change phonologically over time (diachronic phonology) is a central topic in historical linguistics (Campbell, 1998). The ques- tions involved range from reconstruction of ancient word forms, to the elucidation of phonological drift processes, to the determination of phylogenetic re- lationships between languages. However, this prob- lem has received relatively little attention from the computational community. What work there is has focused on the reconstruction of phylogenies on the basis of a Boolean matrix indicating the properties of words in different languages (Gray and Atkinson, 2003; Evans et al., 2004; Ringe et al., 2002; Nakhleh et al., 2005). In this paper, we present a novel framework, along with a concrete model and experiments, for the prob- abilistic modeling of diachronic phonology. We fo- cus on the case where the words are etymological cognates across languages, e.g. French faire and Spanish hacer from Latin facere (to do). Given this information as input, we learn a model acting at the level of individual phoneme sequences, which can be used for reconstruction and prediction, Our model is fully generative, and can be used to reason about a variety of types of information. For exam- ple, we can observe a word in one or more modern languages, say French and Spanish, and query the corresponding word form in another language, say Italian. This kind of lexicon-filling has applications in machine translation. Alternatively, we can also reconstruct ancestral word forms or inspect the rules learned along each branch of a phylogeny to identify salient patterns. Finally, the model can be used as a building block in a system for inferring the topology of phylogenetic trees. We discuss all of these cases further in Section 4. The contributions of this paper are threefold. First, the approach to modeling language change at the phoneme sequence level is new, as is the spe- cific model we present. Second, we compiled a new corpus1 and developed a methodology for quantita- tively evaluating such approaches. Finally, we de- scribe an efficient inference algorithm for our model and empirically study its performance. 1.1 Previous work While our word-level model of phonological change is new, there have been several computational inves- tigations into …

Content truncated to 500 words