A Lexical Simplification Framework for Turkish

Lexical simplification is a fundamental step towards improving the accessibility, comprehension, and readability of texts, particularly in languages with limited linguistic resources. In this study, we adopt a state-of-the-art lexical simplification approach and propose a lexical simplification framework tailored to Turkish. Our framework leverages a combination of complex word identification tasks and substitution generation through pre-trained language models to identify complex lexical units using selective substitution ranking approaches and algorithms and replace them with simpler alternatives, thereby improving text readability. This work makes three key contributions: (i) a comprehensive study of lexical simplification for Turkish, including the complex word identification subtask; (ii) a rigorous comparison of various language models for candidate generation using the masked language modeling objective; and (iii) an in-depth exploration of the impact of different complexity thresholds and additional parameters on overall performance. Our framework demonstrates a strong capability to balance simplification and contextual preservation, offering an effective solution to lexical simplification in Turkish.