When you think technology, you think fast, agile, and most likely young. World leaders in the field of computers, coding and software have left an indelible mark on mankind in the last, but largely, four decades – Microsoft was founded in 1975, Apple in 1976, Amazon in 1994, Google in 1998… But there’s only one company with a century-old legacy that’s still highly relevant, developing cutting-edge technologies and a world leader in patents. That’s IBM.

Founded in 1911, the technology major has remained a powerhouse for 110 years now. The company has not only witnessed history being made, but also been a part of some incredible world events – aiding the historic Apollo Missions, opening the doors to nanoscience, developing a conversational AI platform that has won the Jeopardy! Challenge, and beating grandmaster Garry Kasparov. These momentous milestones would not have been possible without IBM Research – considered the innovation engine for IBM, and setting the precedent for fostering scientific creativity in companies – leading the way from programmable computers to quantum computers.

In India, IBM Research continues to reflect these tenets of boundless innovation. Established in 1998, the division is a premier research lab in the region, and spans across New Delhi and Bangalore. Karthik Sankaranarayan leads the Interactive AI Lab in Bangalore, managing a team of 26 engineers and scientists, and driving some of IBM India’s most advanced projects in Natural Language Processing (NLP). After completing his PhD in 2011 from Ohio State University in Artificial Intelligence, specifically Computer Vision and Machine Learning, Sankaranarayanan joined IBM. He has worked on several projects including Project Debater, Question-Answering, and developed a range of AI applications for financial services, telecom, healthcare and agriculture. A typical work week, he describes, is a combination of working towards a scientific paper deadline, brainstorming on patents and collaborating with product teams to understand customer challenges. “Not only do we brainstorm today’s problems, but work on addressing future challenges too.”

(Karthik Sankaranarayanan, Lead, Interactive AI - IBM Research India)

At IBM, the endeavour is to help large corporations, businesses and enterprises make sense of the tremendous volumes of information they possess, to serve their customers better. “NLP doesn’t just learn and analyse language – it also delves into interactions taking place in these domains between customers, and extracts insights by analyzing sentiments, intonations, dialects and more. This will enable AI models aid in accurate decision making,” he explains.

For IBM, there are three core research pillars within NLP: Interactive AI systems; AI for Indian languages and Making AI trustable.

Interactive AI systems dates back to the initial days of the Watson platform and its efficacy in domains like healthcare and finance. The technology has matured greatly from free-form Q&A to complex conversations and argumentative structures such as debates, discussions on policy, economic developments and more. Trustable AI, as the name suggests, is a pressing need to make the very technology IBM is priming, more reliable, free of bias, objective and fair.

But one of the most exciting pillars for IBM is AI for Indian languages, where multiple advancements focused on Indian languages are being witnessed. India is a key market for IBM, keen on democratizing Indian languages for the Internet. Less than 10% of Indians speak and transact in English. The remaining 90% is split between multiple languages like Hindi, Bangla, Tamil, Telugu, Gujarati and more. While the opportunity is hugely exciting, its also challenging – scripts are different from English, pure datasets for regional languages are much harder to come by. In addition, India’s multilingual tendencies lead to mixed language sentences, bringing about another complexity of code-switching. “For instance, my niece once said: “Lift-oda button press pannu” which in English-Tamil means “Press the lift button”. Being a native Tamil speaker who is also proficient in English, I understood what she meant. But, will a computer understand that sentence? Likely not. And these overlaps are more common than we think – be it at a workplace or at home,” says Sankaranarayanan.

Currently, given the share of Hindi speakers in India is about 50%, IBM is working on tuning the Watson platform in Hindi. This involves reading and understanding complex documents in Devnagri script, with the help of basic NLP, advanced NLP and sentiment analysis. This is especially useful for IBM’s clients in financial services, telecom and agriculture. The developments in written Hindi are paving the way for the teams to work on spoken Hindi data as well. In addition, they have also started working on other regional language NLP models. Sankaranarayanan envisages computational argumentation to be the next major challenge for sentiment analysis – a complex yet popular tool for organisations today. Building computational models for arguments, supporting them with authentic claims and evidence will be the near term challenges that engineers will be working on. And for a market like India, there are linguistic and cultural variations to consider too: “Idioms, metaphors and phrases are still complex for systems to understand. For instance, consider the popular Hindi idiom “behti Ganga mein haat dhona” – it literally translates to washing your hands in the river Ganges – and that’s how the computer will understand it! The machine needs to be taught these regional nuances better to understand what the customer is trying to convey when he speaks or types such sentences.”

While the world came to a grinding halt last year due to the COVID19 pandemic, IBM Researchers were doubling down on their creativity. One of the most interesting projects the team was engaged in was delivering a rich, virtual viewing experience of the US Open. The pandemic had forced the world’s biggest sporting events to be viewed remotely, and this meant the new, virtual experience had to be that much richer and immersive. Sankaranarayanan and the IBM Research team in Bangalore worked on developing natural language generation technology which allowed viewers to access pre-game information, data and game statistics on players, their previous games and more. This led to a significant bump in online traffic.

However, they also got an opportunity to work with ICMR to develop a Watson chatbot to coordinate with diagnostic labs and assist them with COVID19 screening and testing. The portal was especially helpful in ramping up testing across remote locations with language barriers. The success with ICMR led IBM to develop a similar system for Andhra Pradesh government to dispense vital and timely information to citizens. In addition to healthcare, IBM Research sees a lot of opportunity in driving NLP in education such as developing quality content in local language and assisting teachers in grading & evaluation.

The growing synergy between research, academia, industry and government is a harbinger of innovation in India, and IBM Research is capitalizing this newfound brain-trust to nurture scientific thinking. Sankaranarayan says, “Most of us in R&D come with deep roots to academia, and we extend those relationships into the professional realm. We have struck a strategic partnership with IIT Bombay and IIT Delhi, as part of the AI Horizons Network, where top universities and research institutions are connected to IBM through multiple collaborative projects in areas like NLP, speech recognition, core ML and knowledge graphs. It allows us a chance to work with young students and foster a culture of innovation early on.”

 

Want to publish your content?

Publish an article and share your insights to the world.

ALSO EXPLORE

DISCLAIMER

The information provided on this page has been procured through secondary sources. In case you would like to suggest any update, please write to us at support.ai@mail.nasscom.in