A Statistical Model for Lost Language Decipherment

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 1048–1057, Uppsala, Sweden, 11-16 July 2010. c⃝2010 Association for Computational Linguistics A Statistical Model for Lost Language Decipherment Benjamin Snyder and Regina Barzilay CSAIL Massachusetts Institute of Technology {bsnyder,regina}@csail.mit.edu Kevin Knight ISI University of Southern California knight@isi.edu Abstract In this paper we propose a method for the automatic decipherment of lost languages. Given a non-parallel corpus in a known re- lated language, our model produces both alphabetic mappings and translations of words into their corresponding cognates. We employ a non-parametric Bayesian framework to simultaneously capture both low-level character mappings and high- level morphemic correspondences. This formulation enables us to encode some of the linguistic intuitions that have guided human decipherers. When applied to the ancient Semitic language Ugaritic, the model correctly maps 29 of 30 letters to their Hebrew counterparts, and deduces the correct Hebrew cognate for 60% of the Ugaritic words which have cognates in Hebrew. 1 Introduction Dozens of lost languages have been deciphered by humans in the last two centuries. In each case, the decipherment has been considered a ma- jor intellectual breakthrough, often the culmina- tion of decades of scholarly efforts. Computers have played no role in the decipherment any of these languages. In fact, skeptics argue that com- puters do not possess the “logic and intuition” re- quired to unravel the mysteries of ancient scripts.1 In this paper, we demonstrate that at least some of this logic and intuition can be successfully mod- eled, allowing computational tools to be used in the decipherment process. 1“Successful archaeological decipherment has turned out to require a synthesis of logic and intuition ...that comput- ers do not (and presumably cannot) possess.” A. Robinson, “Lost Languages: The Enigma of the World’s Undeciphered Scripts” (2002) Our definition of the computational decipher- ment task closely follows the setup typically faced by human decipherers (Robinson, 2002). Our in- put consists of texts in a lost language and a corpus of non-parallel data in a known related language. The decipherment itself involves two related sub- tasks: (i) finding the mapping between alphabets of the known and lost languages, and (ii) translat- ing words in the lost language into corresponding cognates of the known language. While there is no single formula that human de- cipherers have employed, manual efforts have fo- cused on several guiding principles. A common starting point is to compare letter and word fre- quencies between the lost and known languages. In the presence of cognates the correct mapping between the languages will reveal similarities in frequency, both at the character and lexical level. In addition, morphological analysis plays a cru- cial role here, as highly frequent morpheme cor- respondences can be particularly revealing. In fact, these three strands of analysis (character fre- quency, morphology, and lexical frequency) are intertwined throughout the human decipherment process. Partial knowledge of each drives discov- ery in the others. We capture these intuitions in a generative Bayesian model. This model assumes that each …

Content truncated to 500 words