Tracking the evolution of translated documents: revisions, languages and contaminations

Abstract

Dealing with documents that have changed through time requires keeping track of additional metadata, for example the order of the revisions. This small issue explodes in complexity when these documents are translated. Even more complicate is keeping track of the parallel evolution of a document and its translations. The fact that this extra metadata has to be encoded in formal terms in order to be processed by computers has forced us to reflect on issues that are usually overlooked or, at least, not actively discussed and documented: How do I record which document is a translation of which? How do I record that this document is a translation of that specific revision of another document? And what if a certain translation has been created using one or more intermediate translations with no access to the original document? In this paper we addresses all these issues, starting from first principles and incrementally building towards a comprehensive solution. This solution is then distilled in terms of formal concepts (e.g., translation, abstraction levels, comparability, division in parts, addressability) and abstract data structures (e.g., derivation graphs, revisions-alignment tables, source-document tables, source-part tables). The proposed data structures can be seen as a generalization of the classical evolutionary trees (e.g., stemma codicum), extended to take into account the concepts of translation and contamination (i.e., multiple sources). The presented abstract data structures can easily be implemented in any programming language and customized to fit the specific needs of a research project.