Anup Joseph

I am an enthusiastic student looking for opportunities to prove myself.I specialize in the field of Machine Learning.
These are some of the projects I have built as part of my college work and participating in competitions

Using Probabilistic Tag Modeling to Improve Recommendations

• User behaviour on a website is extremely valuable information which helps to identify the similarities among users and help to target a subset of users in a better informed and more informed ways.

• In this project we modelled the information in form of a knowledge graph with
  Nodes -> search term in the websites
  Edges -> clicks and purchases on a product after the search

• After creating such a structure we applied walktrap community detection over the graph to cluster the similar search terms together. The pagerank algorithm was applied over every sub-cluster to find out the central tag in the every sub-cluster. Various algorithms were then applied to provide recommendations. For further information please see this file
  http://ml4ed.cc/attachments/GokkayaUsing.pdf

Automated Land use classification Using AI/ML

• Develop a deep-learning based software for automatically classifying land-use from multi-temporal multi-spectral high-resolution satellite imagery. The developed model should be scalable/efficient to allow rapid mapping of incoming datasets and must incorporate a web-based viewer for visualizing input as well as classified output. The viewer interface must also allow the user to visualize changes that have occurred within a given timeframe.

Semantic Classification of call-center conversations

• We created a tri-layered Dockerized pipeline for analyzing and understanding customer satisfaction in cal-center calls.
• Diarization - Aalto ASR software to diarize i.e. to separate the voices among the call-center employee and customer
• Transcription - Using Mozilla deepspeech model to create a transcript of the audio. Used transfer learning to improve the accuracy of the model on Indian English
  Emotion Analysis - Tried out different approaches to understand the emotion of the customer from customer voice during various periods of the call and assign a composite score at the end of the call