07 August, 2018:
“My interest as a researcher was always to work on public policy application,” said Dr. Nandana Sengupta, faculty, Azim Premji University, at a seminar hosted by IIMB’s Research & Publications Office, on July 20.
Dr. Nandana Sengupta, faculty, Azim Premji University
She initiated her talk by defining informal labourers. “Unorganized workers consist of those working in the unorganized enterprises or households, excluding regular workers with social security benefits, and the workers in the formal sector without any employment or social security benefits provided by the employer,” she explained.The paper being discussed was based on the declared slums of Bangalore. The larger research objectives were to form an informal sector skills map, to match workers to jobs and training, to improve training modules, to automate CV generation and skill certification.
She explained that her study had a twofold objective. First, a bottom-up crowdsourcing approach to map the space of skills, qualities and personality traits among Bangalore’s informal workers, and second, to assess the applicability of machine learning tools like recommendation systems and adaptive surveys to a public policy context in India.
The research for this paper consisted of the survey and the analysis. For the survey, one member from each family, under consideration, was contacted and they were asked the relevant questions. In the survey, “What skills, knowledge or values do you have which you think will help you in getting or keeping a job? List as many as you can think of”, was asked to the candidates.
A section of audience at the seminar
More than 5000 skills and values like riding a bike, speaking languages, etc. were captured and this, in turn, acted as input to the second round of surveys and machine learning techniques. A relevant problem with this question is the fact whether the people in the survey know what a skill is. The analysis of the basic skills by gender, the Mincerian Wage Equation with Hackman correction and several machine learning applications such as LASSO (which is a regression analysis method), matrix factorization and active learning were used.
Using Mincer earnings function which is a single-equation model, Dr. Sengupta explained wage income as a function of schooling and experience. She also explained how matrix factorization helped in matching the skill space.
Inviting questions from the audience, Dr. Sengupta touched on topics like improving training modules and skill certification methods.
