Ab Initio: Automatic Latin Proto-word Reconstruction

Proceedings of the 27th International Conference on Computational Linguistics, pages 1604–1614 Santa Fe, New Mexico, USA, August 20-26, 2018. 1604 Ab Initio: Automatic Latin Proto-word Reconstruction Alina Maria Ciobanu, Liviu P. Dinu Faculty of Mathematics and Computer Science, University of Bucharest Human Language Technologies Research Center, University of Bucharest alina.ciobanu@my.fmi.unibuc.ro, ldinu@fmi.unibuc.ro Abstract Proto-word reconstruction is central to the study of language evolution. It consists of recreating the words in an ancient language from its modern daughter languages. In this paper we inves- tigate automatic word form reconstruction for Latin proto-words. Having modern word forms in multiple Romance languages (French, Italian, Spanish, Portuguese and Romanian), we infer the form of their common Latin ancestors. Our approach relies on the regularities that occurred when the Latin words entered the modern languages. We leverage information from all modern languages, building an ensemble system for proto-word reconstruction. We use conditional ran- dom fields for sequence labeling, but we conduct preliminary experiments with recurrent neural networks as well. We apply our method on multiple datasets, showing that our method improves on previous results, having also the advantage of requiring less input data, which is essential in historical linguistics, where resources are generally scarce. 1 Introduction Proto-language reconstruction is one of the main concerns of historical linguistics and is central to under- standing the evolution of the world’s languages. Reconstructing an ancestral language from the modern languages descending from it not only allows an analysis of language change across space and time, but also reveals information about the historical relationships between languages. The traditional pro- cess of reconstructing ancient languages is called the comparative method (Anttila, 1989) and relies on cognates (words in different languages descending from a common proto-word). The main idea of the comparative method is to perform a property-based comparison on multiple sister languages in order to infer properties of their common ancestor. For a long period, the comparative reconstruction has been a time-consuming manual process that required a large amount of intensive work. The first step of the process consists in identifying cognate pairs. Identifying cognates and borrowings by means of computational approaches has attracted considerable attention in recent years (Hall and Klein, 2010; Tsvetkov et al., 2015; Ciobanu and Dinu, 2015). How- ever, few studies went beyond this step, and beyond the comparative method, to automate the process of proto-language reconstruction (Oakes, 2000; Bouchard-Cˆot´e et al., 2013; Atkinson, 2013). Reconstruct- ing proto-words is a challenging task. While the main hypothesis in this research problem is that there are regularities and patterns in how words evolved from the ancestor language to its modern daughter languages, there are also words which diverged significantly from their ancestor. Take, for example, the Latin word umbilicu(lu)s. It evolved into buric (Romanian), nombril (French), and umbigo (Portuguese). One of the best approaches to proto-word reconstruction (Bouchard-Cˆot´e et al., 2013) relies on an anal- ogy with reconstructing the genealogy of the species from genetic sequences in biology. This approach requires an existing phylogenetic tree and the phonetic transcripts …

Content truncated to 500 words