Community IVR: Voice AI for offline communities (Cal Hacks 8.0)
During COVID-19, Indonesia shared health advice, social support rules, and emergency information mainly through apps and social media. However, an estimated 40% of Indonesians had no smartphone. Rural, older, and low-income people lost access to information while facing some of the greatest risks.
Standard phone menus assume patient users who can read and speak clearly. However, background noise, unfamiliar technology, and local accents disrupt these menus. Indonesian speech recognition trained on formal broadcasts performs poorly on Javanese and Sundanese accents. It also struggles when speakers switch languages. Moreover, quarantine rules, social aid, and emergency contacts changed weekly.
I designed an assistant for basic phones and 2G networks. It required no app or internet service. Moreover, speech recognition handled Indonesian with Javanese and Sundanese accents. A knowledge graph stored connections between health advice, service rules, and emergency contacts. Then text-to-speech read the answers aloud.
I designed the service for people without digital skills. Therefore, the interaction followed familiar banking and telecom phone menus operated with keypad tones. These are called DTMF IVR systems. Moreover, people unfamiliar with menu navigation could ask questions freely by voice.
We curated the knowledge graph for Indonesia. It covered local health contacts, regional quarantine rules, and eligibility for social assistance. Moreover, it included local differences in how people asked questions.
The service addressed a gap affecting the estimated 40% of Indonesians without smartphones. Then it won Best Community Track at UC Berkeley’s Cal Hacks. Moreover, its basic-phone, speech, and knowledge-graph design can serve other regions with limited smartphone and internet access.