The Korean Speech Recognition Sentences: A Large Corpus for Evaluating Semantic Context and Language Experience in...

Area of Science:

• Linguistics
• Speech Science
• Psycholinguistics

Background:

• Developing standardized linguistic materials is crucial for speech and language research.
• Korean language resources for speech recognition and lexical processing studies are limited.
• Understanding semantic predictability's role in speech perception is key.

Purpose of the Study:

• To create and validate a large Korean sentence corpus.
• To establish sentences with controlled degrees of semantic predictability (predictable, neutral, anomalous).
• To facilitate research in Korean speech recognition and lexical processing.

Main Methods:

• Generated 1,021 Korean sentences varying in final-word predictability.
• Validated semantic predictability using cloze tests with native (n=56) and nonnative (n=19) Korean speakers.
• Employed a computer language model for an efficient, comparable predictability evaluation.

Main Results:

• Speech-in-noise experiments showed sentence intelligibility comparable to English corpora.
• Intelligibility significantly varied based on semantic predictability conditions.
• Sentences demonstrated appropriate difficulty for assessing effects of noise and language experience.

Conclusions:

• The validated Korean sentence corpus is a valuable resource for speech research.
• Enables investigation of speech perception in Korean across diverse conditions.
• Contributes to expanding linguistic corpora for international speech research.