Tech
Spoken Turkish into sign language, offline and in real time
Doğa Ülkü is building ARIA, a voice-to-sign-language system that converts spoken Turkish into Turkish Sign Language in real time, with an accuracy she puts at 78%.
"Every existing Turkish Sign Language app requires you to type," Doğa Ülkü says. "Nobody speaks to it. Nobody built emotion into it." So she analysed 596 real user reviews across four apps, found the failure points and built something different.
ARIA, the Accessible Real-time Interpretation Assistant, converts spoken Turkish into Turkish Sign Language in real time using offline AI. No internet is required. It detects emotion through three layers combined: audio, text, and rules. She puts its accuracy at 78%, and says it is the first system in Turkey to do this. It is aimed at Turkey's deaf and hard-of-hearing population, which she counts at 3.6 million people. She started it, she says, "because I kept seeing a gap nobody was fixing."
“I'm not waiting to feel ready. I'm already building.”
Doğa Ülkü
She studied Industrial Engineering at Manisa Celal Bayar University, graduating in June 2026. She co-authored a peer-reviewed paper on multimodal AI published at an international symposium, was recognised as Microsoft Young Technology Star 2024, and has won competitions judged by NASA.
ARIA, she says, is not a competition project but "the one I keep coming back to at midnight because it actually matters."