Empirical Investigations Of Generative Ai İn Expetrimental Linguistics The Estimation Of Word Metrics

Experimental linguistics requires accurate word metrics since word processing is influenced by frequency, familiarity, imageability, and concreteness. Obtaining accurate metrics, however, can be challenging due to the time-consuming and costly nature of data collection, making word metric values unavailable for many languages. Although metrics exist for some Indo-European languages such as English, it is not practical or even possible to carry out a study to estimate word metrics for all possible words in a language. Recent developments in Generative Artificial Intelligence (GenAI) offer a viable solution to this problem, as these models can estimate word metrics for different languages and contexts. Nevertheless, the accuracy of these estimates should be investigated thoroughly, as unreliable estimates can lead to misinterpretations of how people process language. Using an experimental approach, this book investigates whether GenAI can provide accurate estimates for word frequency, familiarity, imageability, and concreteness. The chapters focus on Turkish words to examine whether popular GenAI models can be reliable tools for estimating word metrics in a non-Indo-European language. The results are mixed, as some AI-estimated word metrics were nearly identical to human ratings, whereas others differed from the human baseline, even when reference values were provided. This book offers a series of systematic investigations for researchers and instructors looking to use AI models in estimating word metrics reliably.