The Wahoo.ai team, consisting of three final-year informatics engineering students - Hafid Abi Daniswara, Aulia Adila, and Fabian Savero Diaz Pranoto, clinched the top spot in the Tools Generative AI Development category. This competition was held as part of the Artificial Intelligence Innovation Summit (AIIS) 2023 by the Collaboration on Artificial Intelligence Research and Industry Innovation (Korika), a research institution focusing on AI development for creative and revolutionary solutions.

The trio proposed the development of an Augmentative and Alternative Communication (AAC) application for individuals with aphasia, a common condition following a stroke that affects speech ability. Aphasia patients often face challenges in articulating complete or intelligible speech, prompting the need for a generative AI approach.

They named their web-based application GemaKata, designed to transform speech from aphasia sufferers into a transcript using the speech-to-text feature. The transcript undergoes correction and completion through ChatGPT's sentence completion method, making it more understandable for the listener. Finally, the transcript is retransformed into voice through the text-to-speech feature.

Aulia expressed hope, stating, "After the completion of an aphasia patient's sentence, the listener could understand the real meaning of the speech more easily."

Hafid, Fabian, and Aulia emphasized extensive research in the medical field before implementing their generative AI for aphasia patients. They observed a relatively limited application of generative AI in the medical field compared to other areas, recognizing the promising potential of their proposed solution.

Fabian highlighted, "The biggest challenge was transforming voice to text and predicting the produced sentence. Especially the prediction, we had to find the right prompt or instruction so the AI could produce the desired output.

"Despite the successful application, the team acknowledges room for improvement in the GemaKata application. Enhancements in the quality and processing speed of both the speech-to-text and text-to-speech features are considered. They also plan to develop their own model to customize output for aphasia sufferers and address scalability issues.

Reporter: Hanifa Juliana (Urban and Regional Planning, 2020)

Editor: M. Naufal Hafizh