​Research: MidcurveNN: Encoder-Decoder Neural Network for Computing Midcurve of a Thin Polygon

Various applications need lower dimensional representation of shapes. Midcurve is one-dimensional(1D) representation of a two-dimensional(2D) planar shape. MidcurveNN, a novel method, uses Encoder-Decoder neural network for computing midcurve from images of 2D thin polygons in supervised learning manner.

<Read More> <ODSC Presentation>


Legal Informatics

Information Retrieval competition at FIRE 2017 IRLeD track. Submission was done as Right Steps team.

  • Catchphrase Extraction

           1st prize  winning entry, a CRF based solution was submitted at the competition... <See Results>

  • Precedence Retrieval

           8th ranked entry, a Regex+Doc2Vec based solution was submitted at the competition... <See Results>

GST Chatbot

A chat bot for queries related to Goods and Services Tax (GST) in India. This is a beginning (primitive/simplistic) version to be built upon further to be practically useful. <Read Article> <Watch Video>

Lawyer Assistant

With thousands of court cases to refer, a lawyer would need a smart assistant to extract highly relevant information about the case s/he is building. Here is a PoC of such NLP based solution... <Watch Video>
Salient features:

  • Supreme court judgements are scraped from public website.
  • Given a theme, related judgements are searched based on topics
  • Given a judgement, its similar judgements are fetched
  • For a selected judgements, its keywords are extracted and displayed
  • Its full text and computed summary is also presented.Type your paragraph here.