Worked under Ankur Pandey on 

1) Automatic timeline Generation from news articles on a particular events, person etc (Eg: CPEC,  WORD TRADE ORGANIZATION) using various Machine learning techniques.
  • Wrote heuristic and regex functions for Date extraction using published date of news articles.
  • Used Language model to get importance of the events. 
  • Used hierarchical  cluster to remove duplicate events.
  • Use technique similar to KDE to finally create the timeline. 

2) Improved Headline clustering and merging algorithms:
•  Used Inverted index to reduce time complexity of clustering from 15-20 min to 8-13 min and for merging from 8-10 min to 30-45 sec.
• Used Hierarchical clustering with better distance matrix and tuned parameters to get better clustering results.

Worked under Abhijit Bhole to develop and implement
1) A probabilistic language model for intent matching in sales call.
2) Simple search engine using bag of words and KL divergence.

SDE at MindTickle

Abhimanyu Ray

SDE at MindTickle