TYPOLOGICAL DIFFERENCES BETWEEN SYNTHETIC AND ANALYTICAL CAUSAL STRUCTURES IN THE KAZAKH LANGUAGE

Abstract

This article presents a methodology for the manual annotation and structural-semantic systematization of causal constructions in the Kazakh language based on the KazCausal annotation scheme. The aim of the study is to formally describe the morphological, lexical, and syntactic means of expressing causality in Kazakh through an annotation model and to provide a scientific basis for the development of a corpus adapted to natural language processing tasks. The study is based on 1,250 causal sentences selected from scientific, journalistic, and normative texts. All sentences were considered as linguistic units expressing cause-and-effect relations and were classified into three main structural types: SYNTHETIC_CAUSE, ANALYTICO_SYNTHETIC_CAUSE, and ANALYTIC_CAUSE. The results show that synthetic constructions account for 431 sentences (34.5%), analytico-synthetic constructions for 425 sentences (34.0%), and analytic constructions for 394 sentences (31.5%). The proposed annotation scheme makes it possible to simultaneously annotate the cause fragment, the effect fragment, the causal marker, the structural type, the specific morphological form, and the semantic subtype. The quality of annotation was evaluated using an F1 score of 0.81 and Cohen’s kappa coefficient of k=0.72. The findings of the study provide a foundation for the development of typologically adapted NLP resources aimed at the automatic detection of cause-and-effect relations in the Kazakh language.