Compatibility of logical semantic relations: Methods of quantitative analysis

The paper deals with logical semantic relations (LSR) that ensure text coherence and examines their explicit markers in text, i. e., connectives. An overview of existing approaches to definition and classification of LSRs is presented; different LSR combination patterns are examined. To make quantitative assessment of LSR compatibility, one should be able to annotate LSRs and their markers in texts. To support such annotation, a supracorpora database was developed, including customizable faceted classifications for LSRs and connectives. Based on the database data, quantitative information on relations combinations was accumulated and interpreted. For example, it was demonstrated that adjoining relations collocate with extensional generalization relations much more often than with relations of specifications; discourse realizations commonly used to mark more than one relation at a time were identified, etc. The results of LSR compatibility analysis were used to elaborate the methodology of reversible generalization of information objects. High flexibility and accessibility of the proposed approach allows researchers to address the underinvestigated problem of LSR compatibility.