Implicit logical-semantic relations in parallel texts: Annotation principles

The problem of implicit logical-semantic relations (LSRs) annotation is considered. The state-of-the-art in the annotation of implicit LSRs is analyzed. The approaches focused on ($i$) analysis of the global discourse structure; ($ii$) analysis of the local discourse structure; and ($iii$) unification of the data annotated within different frameworks and development of a unified annotation standard are presented. The principles for annotating implicit LSRs in parallel texts are proposed, i. e., target of annotation is a translated correspondence (a pair of fragments from the original and translated texts). Translation correspondences illustrating implicit–explicit mismatch have been studied, i. e., where LSR markers are absent in the Russian text while in the text in another language, on the contrary, they are present. Taking into account the specificity of implicit LSRs, the following principles of their annotation were formulated: ($i$) it is necessary to determine the boundaries of LSR arguments (to ensure clarity and convenience of analysis); ($ii$) features of text blocks should form a hierarchical structure (to ensure convenience of using a large number of features); and ($iii$) if a feature of a text block has a lexical marker, this marker should be indicated (to ensure better justification of the annotator's decisions).