Methods for retrieval of implicit logical-semantic relations from parallel texts

The paper focuses on the problem of targeted retrieval of fragments with implicit logical-semantic relations (LSRs) from parallel texts. In order to answer the question whether such retrieval is possible, the author considers the potential of parallel text as a source of new linguistic knowledge and reviews the works that report results of implicit LSRs studies. The aim of the survey is to identify the methods used to create arrays of text fragments with implicit LSRs. It is revealed that parallel texts offer extensive opportunities for retrieval of such fragments, owing to the fact that an explicit LSR in language $A$ may signal an implicit LSR in language $B$ (that is called a bilingual explicit/implicit mismatch). The following methods have been successfully applied for detection of such mismatches: ($i$) discourse parsing; ($ii$) reverse machine translation; and ($iii$) bilingual exclusion search. Therefore, it can be argued that parallel texts — unlike the monolingual ones — can solve the task of targeted retrieval of fragments with implicit LSRs.