Jun 2022 – Apr 2025 · 2 yrs 10 mos

Mar 2019 – May 2022 · 3 yrs 2 mos

Nov 2016 – Feb 2019 · 2 yrs 3 mos

• I am responsible for the designing and development of NLP based solutions, currently working in the domains of finance and procurement.

Sep 2015 – Mar 2016 · 6 mos · New Delhi Area, India

• Evaluation of state-of-the-art subgraph extraction algorithms for comprehendibilty: The project is based on evaluating the graph snippets for interestingness and comprehendibility.
• Extract graph snippets from large entity relationship graphs, such as DBpedia, Yago, and Espresso, given an entity pair using state-of-the-art subgraph extraction algorithms.
• Prepare Evaluation Portal to obtain user preferences of graph snippets.

May 2014 – Jul 2014 · 2 mos · Bangalore Area, India

• Predicting Future Activites of User on Social Media Platforms: The project aimed at predicting future activities of a social media user; we experiment with twitter user.
• Consider twitter stream of a user as time-series data and predict features such as number of tweets, topics, etc, for time instance.
• Propose a time series model based on regression and ARIMA models to predict such features. Compare with other models such as ARIMA, VARMA and DRM.
• The patent application for the work has been filed.

Jul 2011 – Jan 2018 · 6 yrs 6 mos · New Delhi Area, India

• Worked on the problem of Entity-centric Summarization that aims to generate comprehendable text summaries of graph snippets. These graph snippets or subgraphs are obtained for a set of entities from large entity-relationship graphs. The problem is divided into following subproblems:
• Evaluation of the state-of-the-art subgraph extraction algorithms for comprehendibilty.
• Build system to generate text summaries for subgraphs.

May 2010 – Jul 2010 · 2 mos · Noida Area, India

• Text Event Parser : Event is a social occasion with attributes like date, time, venue, contact details,etc. Given a domain url, it extracts all events and associated details.
• Designed a single schema comprising of the various details (such as date, time, title, description, contact information) available from different event related sites.
• Developed a system based on rule based approach to extract the events and associated details. For each feature, a different extraction technique was used. For example, for date and time, python datetime module was modified to extract all expected formats and to identify title, formatting features were used.
• Tagged the events with their types e.g. Music, Spritual, Night-out, etc using dictionary overlap and frequency based technique.