drop me a mail at:
Recent Experience / Projects
INDIAN STATISTICAL INSTITUTE
|Summer 2020||Kolkata, India|
Worked with Prof. Ujjwal Maulik on Non Linear Dimensionality Reduction- Algorithms based on canonical feature extraction techniques such as PCA and LDA applied through Graph Embedding. Presented a short presentation and report on the same to Prof. Sanghamitra Bandyopadhya (Director, ISI Kolkata)
UNSUPERVISED KEYWORD EXTRACTION
|Natural Language Processing||January – March, 2021|
Formed an unsupervised, graph-based algorithm for Keyphrase Extraction that exploits syntactic relations using dependency parsing, augmented with local text attributes. We reconstruct dependency trees in a Hyperbolic space to locate key information in a document, and then rank candidates on statistical features. To our knowledge, this is the first work that uses hyperbolic geometry for keyphrase extraction.
REFERENCE MANAGEMENT TOOL
|Semester long project||September – November, 2020|
A Dashboard created with MySQL and Python backend to collect Metadata of Research Papers, allowing user to effectively search, edit and organize. Topic modeling was implemented using LDA to group papers. The tool took input as title or PDF and automatically retrieved Metadata by scraping through Google Scholar and CrossRef.
MODELING MOVEMENT OF AQUATIC ECOSYSTEMS
|Mathematical Modeling||March, 2020|
Modeling the movement of Pelagic fish stocks in North East Atlantic ocean for the next fifty years: Fisheries face the need to relocate themselves over time due to global warming, as fishes move away to colder temperatures. This model predicted their movement by analysing the rise in global and local temperatures as an effect of global warming and suggested relocation for small fisheries accordingly.
SUDOKUS AND GRAPH THEORY
|Semester long project||February – March , 2020|
Interpreting a Sudoku as an NP Graph coloring problem, working through the Chromatic Polynomial and general Combinatorics to analyse the number of Sudoku puzzles and solutions that can be generated with varying number of clues.