Alex Perelygin

M.S. in CS

Department of Computer Science

Stanford University

aperelyg at stanford dot edu


Bio

I received my masters in Computer Science at Stanford University, concentrating in AI. I did my undergrad in Simon Fraser University in Burnaby, British Columbia, jointly in Mathematics and Computing Science.

Research

My research has been focused on the application of deep learning to NLP, though I previously did some work on applying machine learning to education analytics as well. Some projects I worked on are listed below:

Transition-based Dependency Parsing with Deep Learning

I worked on an attempt to apply deep learning to transition-based dependency parsing

Semantic Compositionality and Deep Learning for Sentiment Analysis

We tackled the problem of sentiment analysis with a model aimed at building up a semantic representation of sentences recursively through their parse trees. This appraoch used deep learning to generate these vector-space semantic representations and then classify them into sentiment categories. We used a newly-created sentiment treebank with fine grained annotation of all parse-tree nodes of a movie reviews data set to train our model, which was then able to classify sentiment at all nodes of the parse tree. More information, including a live demo, can be found here.

Learning Analytics

I worked with the Transformative Learning Technologies Laboratory on both vision and natural language analysis of student learning patterns in design. I worked on identifying key characteristics and comparative evaluation of progressive hand-drawn designs, and then later upon assessing student achievement in introductory programming courses (Stanford's CS106A in particular) through analysis of the natural language within the comments of the source code.