ETRI releases Korean database AI KorBERT

A Korean language model or database for artificial intelligence has been released to help understand and get a proper answer during conversation. The new model has been assessed as more proficient in Q&As and object recognition in sentences than the existing Google Korean model.

The Electronics and Telecommunications Research Institute (ETRI) released KorBERT, a deep learning linguistic model specialized in the Korean language, which was independently developed to innovate AI services, the institute announced on Tuesday on its website.

A linguistic model is a kind of a database devised for AI to learn languages. Linguistic values are expressed as figures so that AI can deep-learn them. The system studies the frequencies of words, which are calculated into probabilities, in certain conditions and aggregates lingual data. The ETRI-developed KorBERT is characterized by its way of learning the Korean language --studying news articles and encyclopedias released for the past 10 years.


ashilla@donga.com