A groundbreaking initiative is underway to push the boundaries of Bengali speech recognition technology, with a focus on capturing the rich tapestry of regional dialects. Bengali.AI, a trailblazing community championing research and innovation in the Bengali language, has joined forces with the Islamic University of Technology Computer Society (IUTCS) to host a Datathon that delves into the history and evolution of Bengali speech recognition, shining a spotlight on the diverse regional nuances.
The core objective of this Datathon is to develop a cutting-edge system capable of transcribing Bengali speech across various regional dialects. Bengali.AI has provided a unique speech corpus for the competition, comprising spontaneous speech samples from 373 individuals hailing from ten distinct geographical locations, including Rangpur, Kishoreganj, Narail, Chattogram, and others. With a cumulative length of 80 hours, this invaluable corpus presents an unparalleled opportunity to enhance Bengali speech recognition technology within the realm of regional speech domains.
Significantly, the submissions to this Datathon will contribute to the development of open-source speech recognition methods for Bengali, fostering a collaborative environment for language technology advancement. The online round of the competition will commence on April 1st, 2024, and conclude on April 24th, with the final round scheduled for April 27th. Hosted on the renowned Kaggle platform, participants can engage in this challenge individually or form teams of up to three members, with both undergraduate and graduate students, as well as working professionals, eligible to participate. International teams are welcomed, provided at least one member is Bangladeshi, and team formations across universities are also encouraged.
Bengali.AI, a voluntary initiative driving advancements in Bengali linguistics, has previously partnered with Google in 2022 to develop a standardized Bangla model. Currently, their efforts are focused on capturing the nuances of regional Bengali dialects, a task that requires extensive data and collaboration. Through their remarkable efforts, they have gathered data samples on regional dialects from 27,000 individuals across diverse regions, amassing approximately 100 hours of data for local dialect models. Their ultimate goal is to develop machine learning models capable of accurately predicting regional dialects in Bengali, preserving the rich linguistic heritage of the language.
Copyright©dhaka.ai
tags: Artificial Intelligence, Ai, Dhaka Ai, Ai In Bangladesh, Ai In Dhaka, Bengali.AI, IUTCS
