Along with the new Gemini Storybook feature, Google announced a range of new partnerships and projects aimed at increasing AI language inclusivity.
To be exact, these initiatives include Project SEALD (in collaboration with Aquarium), Project Vaani, and a partnership with Japan’s Yoshimoto Group. These projects are based in a variety of regions in Asia, though for Filipinos, Project SEALD is no doubt the most intriguing of the three.
While many existing AI tools have advanced English processing, this isn’t exactly the case with local languages, especially in a country as linguistically diverse as the Philippines. As such, Google’s Project SEALD (Southeast Asia Languages in AI Development), along with its open-source counterpart called Aquarium, aims to craft a comprehensive online repository of language data from across the region.
Given that there are over 1,000 languages in SEA, Google is partnering with organizations like AI Singapore to develop a central “Aquarium” of Southeast Asian language datasets. With this, local researchers and developers should get easy access to language information for the creation of AI tools. According to Google, this can lead to the following in the Philippines:
- More sophisticated applications capable of understanding Tagalog, Cebuano, Ilocano, and other local languages. This includes voice assistants that comprehend local idioms and translation tools that accurately convey contextual meaning.
- Improved access to information and services in users’ preferred languages.
- New opportunities for Filipino technology professionals to develop AI solutions specifically tailored for local communities.
Meanwhile, Project Vaani is an initiative in India where Google and the Indian Institute of Science are developing a large, open-source collection of speech data from various districts in the country. Currently, the project has gotten over 28,000 hours of audio data, with a target of 156,000 hours – fitting given the sheer number of languages in India alone.
Finally, Google is collaborating with Japan’s largest entertainment agency, Yoshimoto Group. This collaboration aims to globalize Japanese comedy through the AI translation system, “CHAD 2.” Google mentioned that this system has a 90% accuracy in transcribing and translating comedic content, including proper understanding of comedic timing.
While these two Google language projects may not be connected to AI tech’s future in the Philippines like Project SEALD, they’re still impressive initiatives. After all, they can give developers more insight on how AI tech can improve further in the future to support more languages.