Bangalore AI #4 — NLP for blue-collar workers & building data science products that work
“How do we build an assistant that could help people whose first language isn’t English?” — This is the challenge that Madhav Krishna, co-founder and CEO of vahan.ai shared with the Bangalore AI community.
Vahan.ai has developed an AI-driven Virtual Assistant that empowers blue-collar workers, such as sales teams in the field. They do so with an assistant on Whatsapp that can answer FAQs about the job, not only in English but in other Indian languages such as Hindi, Kannada and Tamil. Natural Language Processing (NLP) in any Indian language is a hard problem to solve, purely because of the lack of data that is available. Also, people who communicate use informal language, while texting which leads to a high variability. Madhav spoke about how Vahan.ai was tackling these challenges. The first step involved normalizing words in non-canonical format to a canonical format, using a Statistical Machine Translator that they have built. They also use Google Translate APIs in their stack as well as a custom Intent Classification Model to give accurate answers to their users. A few attendees also shared their own stories about the approaches they followed to implement NLP in Indian languages. This is the true power of the Bangalore AI community — sharing challenges and lessons learned to enable everyone apply Artificial Intelligence.
The first Bangalore AI event took place almost a year ago in February 2017. Bangalore is undoubtedly the tech capital of India and has a thriving AI scene with startups and enterprises alike implementing AI and tackling AI challenges in their own way. This was the fourth Bangalore AI event and it was great to see the community brave the famed Bangalore traffic to attend it. This event was sponsored by Noodle.ai — a company that provides Enterprise AI system to manage business operations.
Following Madhav’s illuminating talk, City.AI’s very own Head of Community, Elena Tosheva, spoke to the crowd over video conference, sharing the latest worldwide happenings in AI.
Vinodh Ravindranath, CTO of Belong.co closed Bangalore AI #4 with his talk on “The secret to building data science products that work”. Vinodh started his interactive talk by explaining how the Google PageRank algorithm and how Google News, a product that he helped build, worked. He spoke about the challenges of sifting through hundreds of sources, ranking thousands of newsworthy stories and then providing personalized and relevant news to the user. Belong.co, a predictive outbound hiring solution, looks at the recruitment space in a similar way. They help find relevant people for a particular job — using AI of course! Vinodh spoke about how his team analyses candidates based on universities, previous jobs, growth patterns, common networks and shared skills and then matches them with companies that have employees with a similar profile. The talk delved deep into the technical aspects of how this predictive machine learning algorithm works and it is really informative to watch.
It was a testament to the quality of the talks that we had an engaged audience who stayed even past our scheduled finish. You can watch the talks from this and our past events on the City AI YouTube channel. There you can find interesting talks from all around the world! A huge thanks to our speakers Madhav, Elena and Vinodh, our event sponsor Noodle.ai and our community partner 91springboard for supporting us. And, of course, to our great community which is eager to share and learn together and which makes every event a pleasure to host.
