Speaker
Description
This study investigates word-formation chains, i.e. sequences of derivationally related words in which each derivative serves as the base for the next. Due to their rich morphemic structure, such chains are particularly complex in Slavic languages. Their analysis is relevant for understanding word-formation and semantic potential, morphotactics, and lexicographic criteria for vocabulary inclusion. The research builds on the project Formant Combinatorics in Slovenian (2022–2025), which applied automatic extraction methods to data from the Word-Family Dictionary of the Slovene Language (BSSJ). Abstract suffix-chain patterns derived from this resource were extended to larger lexical datasets. Preliminary findings showed limited success of automatic extraction methods (24% for the most frequent suffix chain), whereas AI achieved higher accuracy (59%). However, performance declined markedly with increasing chain complexity. The present study aims to evaluate the effectiveness of ChatGPT 5.5 in identifying word-formation chains with nominal, verbal and adjectival bases. It focuses on the impact of chain complexity, the role of part-of-speech within chains, and problematic suffixes and zero-formant derivation patterns. Particular challenges include derivatives without overt suffixes and cases where zero-formant nouns are misclassified as unmotivated words. The analysis examines suffix chains of varying complexity and compare AI performance with previous automatic extraction approaches, with the goal of improving both extraction methods and AI prompting strategies.