China unveils its first multilingual, multimodal AI input method for Tibetan

(ECNS) -- Qinghai Normal University unveiled China’s first multilingual, multimodal AI input method for Tibetan on Sunday in northwest China’s Qinghai Province, integrating text and speech input, image recognition and AI-assisted features into a single system.

Developed by the State Key Laboratory of Tibetan Intelligence, the input method is available for both smartphones and computers.

Its speech-input function supports the spoken and written forms of the Ü-Tsang, Amdo and Kham dialects, as well as mixed Chinese-Tibetan-English speech. Its image-recognition function can identify mixed Chinese-Tibetan-English text in photos and screenshots. 

The system also provides automatic correction, text polishing and two-way translation between Tibetan and Chinese, backed by terminology databases covering 23 fields, including education, law and literature.

It can learn users’ frequently used words and expressions and supports both on-device inference and cloud-based processing.

(By Helen Mo & intern Yang Hongran)