Course Projects
Socially Ranked Question Answering
Technology for Search and Social Media, Spring 2007
Cocobean is a question answering engine which answers natural language questions and the answers are improved as user vote for the answers. The social ranking adds value to the initial algorithmic ranking of answers for the particular question. The system also learns rules and patterns continuously on the answers being voted and applies them for future ranking of results. It is an effort to use the participatory knowledge along with computation. The system is designed to mainly answer factual based questions.
Design of Indoor Bicycle Storage System
New Product Development, Fall 2007
The indoor bicycle storage rack was designed primarily for urban dwellers with small living area and who value their bikes. We designed a life size product within a short duration of time and a rigorous schedule. We followed structured process design methodology over the course of design. We did extensive user research to understand the real needs of the user and design an appropriate and useful product. Later we selected three concepts for rapid prototyping and designed and evaluated them. We selected the best concept by means of user testing and developed the product by means of three iterations, slowly improving the functionality of the product. We presented the stunning final life size ( shown in picture) product at the trade show and received good feedback from the audience.
I learned numerous lessons during the course of the project and also learned team dynamics. This project introduced me to the fascinating world of product design.
Musical Me
Tangible user Interfaces, Fall 2007
In
this project we explored innovative ways to generate music and
demonstrated that generating music by means of gestures would be fun
and more intuitive than learning complex musical instruments. We built
two different prototypes to demonstrate generation of music by
gestures. The first one was with the Wii remote and music would be
generated in accordance to the gesture made with the Wii remote. The
other prototype demonstrated generating music by means of hand
movements or gestures.
Play it by Ear
User Interface Design and Development, Spring 2007
http://courses.ischool.berkeley.edu/i213/s07/projects/findar/?home
Play it by ear is a location-aware application for mobile phones which allows users to find out the events happening around them with extreme ease. We designed the prototype with the major focus on user centered design. We followed well defined user centered design framework for our design. From the user needs assessment to the final prototype design, every stage has verified by means of user study and evaluation. The design improved over iterations and as a result the final functional prototype was simple and fun to use. We paid attention to minute details and it payed off very well in the outcome.
This project taught me the core principles of user centered design and user research. It was an exciting experience and journey throughout the course of this project.
Blog Analysis: Trends and Predictions
Applied Natural Language Processing, Fall 2006
In this project we analyzed the trends and patterns by mining the text collected from various user blogs. As blogs are open discussions and opinions expressed by general people, we expected to see a overall trend and pattern. The text was processed by using NLP techniques such astopic categorization and Name Entity recognition. We were able to predict general trends from the limited data which we had and it provided some useful insights into the current trends among the people whose blogs were extracted.
Other Projects
Speech Oriented Voice Message System
This project was conducted by a researcher from Technology for Infrastructure in Emerging Regions, UC Berkeley. This project involved a study of asynchronous messaging device, Combadge designed by MERL. I was primarily involved in conducting user study of the device among a group a illiterate users in a rural area. This provided a great opportunity to study the implications of design and interface for illiterate users. It was an exciting experience to understand how design speaks to the users whether illiterate or literate.
Automatic Grapheme to Phoneme Conversion for Tamil
During a tamil speech recognition project we realized the need for a grapheme to phoneme. Though Tamil is a phonetic language it had many exceptions and a tool to map the graphemes and phonemes proved seemed valuable. We initally started with a rule based approach to this problem and then realized that it could be automated by supervised learning algorithms with minimal training data. Later we also found that a hybrid approach is highly useful for this problem.
Parallel Implementation of Sparse Cholesky Factorization
This project is a parallel implementation of sparse cholesky factorization to solve huge Symmetric Positive Definite (SPD) matrices. We build a beowulf cluster of 8 IBM PC nodes using LAM-MPI and implemented the solver to test the speed and performance. The parallel programming experience for a highly computational problem offered lot of useful lessons.
Fractal Image Compression for Color Images
This project involved compression of color images using fractal based algorithm. I used quad-tree partitioning technique and compressed the color images in YUV space which gave higher compression ratio compared to RGB based compression. This project guided me in learning various compression algorithms and it gave me a rigorous programming experience.