Reversibility and alternativeness of generalization of connectives translations models in parallel texts

The paper considers the task of annotation of Russian connectives and their translations with the use of a supracorpora database (SCDB). The first distinctive feature of the SCDB is that it supports creation of bilingual annotations that include both rubrics of the investigated linguistic items (i. e., connectives, in this case) and rubrics of their translations. The second feature is that the rubrics assigned by the linguists are in fact elements of faceted classifications. Implementation of these rubrics in the SCDB enables alternativeness of generalization of annotations that represent concrete informational entities in the SCDB. As these entities are created, abstract translation models of different generalization levels are produced. These models preserve certain common characteristics (aspects) of the generalizable annotations. The support of faceted classifications in the SCDB makes it possible to conduct multifaceted statistical analysis of annotations and connectives translation models in the SCDB. Furthermore, these statistical data are verifiable since the generated quantitative data provide direct links to lists of corresponding annotations. The main objective of the paper is to describe reversibility and alternativeness of the generalization processes in the SCDB, which provides a basis for conducting multifaceted and verifiable statistical analysis of annotations and connectives translation models in parallel texts.