I did my Bachelor’s in Electronics and joined a consulting firm (Thorogood) right after, where I
worked for about 3.5 years overall. I’m currently a data scientist at Diageo and I’ve completed 8 months here now. I was working as a BI & Analytics consultant for about 2.5 years and had just started taking up team lead and technical lead positions across projects.
Well, when you’re a BI & Analytics consultant, you spend the whole day with data and how to answer business questions with them, so it’s only natural that you start thinking about going further down that route with data science. This combined with my passion for learning, coding and math, led me to start looking for courses that would allow me to pursue all this without compromising on work experience. After forming our capstone team, we started looking for interesting datasets to work on and it was quite serendipitous that we stumbled across the Airbnb review dataset.
Arriving at the problem statement was itself a journey. We had numerical data and we could have gone with a demand forecasting or price prediction route, but instead we decided to take the NLP route because it was an entirely new domain for us. Briefly, the focus of the project was to extract key themes (such as “host”, “decor”, “location”) present across all user reviews and subsequently rate each property per theme based on the review sentiments.
These themes and ratings were then used to personalize the recommendation for a user, by matching the themes in their past reviews to highly-rated properties for those themes. The idea of the personalized recommendation came as we researched about Airbnb’s deployed recommendation engine – we thought “well, they have all this review data and so far it’s the user who has to manually analyze all the reviews, how can we improve this for the user?”
Because of the nature of NLP techniques and the lack of numerical values such as accuracy, there was a lot of manual effort required in validating different phases of the project. We tackled this by employing and comparing results across multiple models. We also ensured that every model we chose or rejected was backed by a valid reason and understanding of the algorithm.
“This project truly opened up many opportunities for me. As a team, we were able to present our project and win the best paper award at the International BAI Conference in IIM-B last year. This journey of writing and presenting a paper allowed us to grow as researchers.”
I have been able to employ these techniques at my workplace, adding value to the organization’s marketing efforts. I have also been able to build on top of the knowledge gained from the capstone experience through self-study of cutting edge NLP algorithms. I now volunteer as a part time NLP researcher with Coronawhy, a non-profit aimed at tackling Covid19 with NLP and ML.
For more such success stories, watch this space. Upskill with Great Learning’s PGP Data Science and Business Analytics Course today.0