Muhammad Ali

Research Intern at NEWT, May 2013 - May 2015
B.S. Computer Science from NUCES, 2011 - 2015

About Me

I graduated with a BS in CS from the National University of Computer & Emerging Sciences, Lahore, and was part of NEWT during my junior and senior years, My research interests revolve around statistical natural language processing, machine learning and recommender systems.

Research Projects

[1] Disease Surveillance using Newspaper Articles (Discontinued)

With Dr. Lakshminarayanan and Dr. Umar Saif

Due to the unavailability of complete and timely reports from hospitals and health facilities in developing regions, there is a need for disease surveillance systems which rely on readily available unconventional data sources such as news media. We are working to build an autonomous system which utilizes local newspaper articles to monitor disease spread at a fine-grained level. The system removes erroneous articles by cross-referencing the information within amongst sources. Each validated piece of information is then tagged with the appropriate date and location. A trained model uses past data to compute the severity of the spread.

[2] Text Politeness Classification using Deep Learning

With Dr. Mehreen Saeed and Dr. Sultan Muhammad Khan Sial

For my final year project, we used the Stanford Politeness Corpus to train a model that will be able to classify text as polite or impolite. We are trying to use more complex machine learning methods than the current state of the art in hopes to achieve better accuracies. Instead of using handcrafted features inspired by politeness theory, we seek to use recursive neural networks to learn word representations. The recursive structure of these neural networks lets us capture the recursive structure of language and hence enables us to learn representations of intermediate phrases too. The basic hypothesis in all of our experimentation is that, like other sentiment analysis tasks, capturing compositionality would yield better results. There are a lot of interesting ways this could be used to investigate social phenomena over the internet.

N.B. I have also regularly served for updating and maintaining the Punjab Government's Dengue Tracking System, more recently as part of the team that developed the Dengue Intensity Prediction Model.