Prime Minister Narendra Modi used a real-time artificial intelligence tool called 'Bhashini', during his speech at Kashi Tamil Sangamam in Varanasi on Sunday. Describing it as a "new beginning", he highlighted its role in simplifying his communication with the public.
PM Modi said, “Today, the use of new technology has taken place here through artificial intelligence. This is a new beginning and hopefully, it makes it easier for me to reach you.”
Bhashini was customised to cater to the Tamil-speaking audience specifically during the event in Varanasi. It functions as an AI-driven language translation system, enabling conversations among individuals who speak various Indian languages.
What is Bhashini?
Bhashini was unveiled during the Digital India Week in Gujarat in 2022, during which PM Narendra Modi launched "Digital India Bhashini." The primary objective of this initiative is to improve internet accessibility and digital services in Indian languages, integrating voice-based features and encouraging the creation of content in various linguistic domains.
It comes within the ambit of the Ministry of Electronics and Information Technology, Government of India.
Functioning as an AI-powered language translation system, it has been touted as breaking language barriers and facilitating conversations among individuals who speak various Indian languages. Accessible via Android and iOS applications, the platform ensures a user-friendly experience.
According to its website, “Bhashini shall act as an orchestrator to unify and align a large diverse network across government, industry, academia, research groups and start-ups to bring all their contributions into an open repository.”
How does it work?
Bhashini operates via the crowdsourcing initiative called Bhasha Daan, which invites volunteers to provide language data, including text, speech, and images in diverse Indian languages. This data is utilised to train AI models and enhance the overall quality of language technologies.
Bhasha Daan includes four distinct categories: Suno India, Likho India, Bolo India, and Dekho India, each serving a unique purpose.
In Suno India, users have the ability to transcribe audio content or verify transcriptions made by others. Moving on to Bolo India, individuals can not only record sentences in their own voices but also authenticate audio recordings submitted by others.
Likho India provides a platform for users to translate given text into their local language, as well as validate translations submitted by others. Lastly, within Dekho India, contributors can play a role by typing visible text, labeling images, and validating images labeled by others.
The platform also includes Anuvaad, a set of AI tools which helps with translation, transcription, transliteration, summarization, and sentiment analysis in Indian languages. It enables users to communicate, access information, and create content in their preferred Indian languages.
What are the potential challenges in its implementation?
One of the primary challenges that Bhashini faces is that of data scarcity, as it depends on crowdsourcing. Currently, the open repository has 1,686 contributions in Suno India, 1,399 in Bolo India, 1,014 in Likho India and 643 in Dekho India. Despite the efforts of Bhasha Daan, if there is still a lack of sufficient and high-quality data for Indian languages, then that could affect the performance and accuracy of AI models and tools.
The other challenge remains that of the digital divide. Many potential users of Bhashini may not have adequate digital skills or means to use the platform effectively. According to the Mobile Gender Gap Report by GSMA, 20 percent of men and 29 percent of women in India still do not own a mobile phone. While 52 percent of Indian males use mobile internet, women lag far behind, with only 31 percent using mobile internet, according to the report.
Among those who own a mobile, 23 percent of Indian men own a basic phone, 4 percent own a feature phone and rest 48 percent own a smartphone. The corresponding figures in case of women are 28 percent, 3 percent and 29 percent, respectively.