Investigating the Application and Evaluation of Distributional Semantics in the Translation of Humanistic Texts: a Case Study

Digital Humanities are persisting ascending and the need for translating humanistic texts using Computer Assisted Translation (CAT) tools demands for a specific investigation both of the available technologies and of the evaluation techniques. Indeed, humanistic texts can present deep differences from texts that are usually translated with CAT tools, due to complex interpretative issues, the request of heavy rephrasing, and the addition of explicative parts in order to make the translation fully comprehensible to readers and, also, stylistically pleasant to read. In addition, these texts are often written in peculiar languages for which no linguistic analysis tool can be available. We faced this situation in the context of the project for the translation of the Babylonian Talmud from Ancient Hebrew and Aramaic into Italian. In this paper we describe a work in progress on the application of distributional semantics to the informing of the Translation Memory, and on the evaluation issues arising from its assessment.