[4K5-GS-6c] Language media processing

4:15 PM - 4:30 PM JST(7:15 AM - 7:30 AM UTC)

[4K5-GS-6c-04] Quantification of Human Preferences for Orthographic Variants Using Surprisal

〇Misaki Yamaguchi¹, Masato Mita¹, Shinnosuke Isono², Yohei Oseki^1,3 (1. The University of Tokyo, 2. National Institute for Japanese Language and Linguistics, 3. National Institute of Informatics)

Keywords:

orthography,preference,surprisal

Surprisal estimates derived from language models have been shown to index processing load during human reading comprehension. Nevertheless, the extent to which surprisal can explain higher-order cognitive processes, such as preference judgments, remains underexplored. Against this background, the present study aims to assess how well surprisal predicts human preference judgments. Specifically, we manipulate the orthographic realization of the same lexical item in Japanese (kanji, hiragana, and katakana) and examine whether differences in orthographic preference can be quantitatively accounted for by surprisal. The results of the present survey suggest that surprisal measures correspond, to a certain degree, to observed differences in orthographic preference, indicating that surprisal can explain part of the variability in these judgments.