Bookmark and Share
|
  • Text +
  • Text -

Computer Science

Click to watch Chris Fincher '12 discuss his research project.

Browsing Collectable Items to Motivate Continued Engagement in Learning Systems

Alexis Chuck ('12); Mentor: Caitlin Kelleher*

Abstract:  Despite the ubiquitous presence of computers and technology, limited opportunities to learn computing skills result in only a small, homogenous segment of the population entering the field. Looking Glass is a programming environment designed to provide the opportunity for young learners to independently teach themselves computer programming. We hope to promote long-term engagement with Looking Glass by learning from the example of online services such as NeoPets and Webkinz. Young users spend hundreds of hours on these websites in the pursuit of new, desirable, and rare items to add to their collections. In Looking Glass, the corresponding collectables are 3D models of environments, characters, and props used to create an animated movie. We designed a system to employ users’ natural acquisitiveness to motivate sustained interest in Looking Glass. We describe the results of an early formative study on an interface for browsing 3D models that supports this system.
Funding Provided by: Computer Research Association's Committee on the Status of Women - Distributed Research Experiences for Undergraduates *Washington University in St. Louis National Science Foundation Grant #0835438 (CK)

A Linear Regression Model For Monolingual Text Alignment

Will Coster ('13); Mentor: David Kauchak

Abstract:  Text simplification is the process of altering a document to reduce reading complexity by incorporating more accessible vocabulary and sentence structure.  Many text simplification systems train models based on aligned pairs of normal and simple sentence pairs.  Previous approaches align similar sentences found in English and Simple English Wikipedia, but result in modest accuracy (~90%).  In this work, we extend these prior approaches and employ a logistic regression model to predict whether a pair of sentences should be aligned. Using this new model to filter prior corpora we achieve a 3% increase in precision with no loss in recall.
Funding Provided by: Paul K. Richter and Evalyn E. Cook Richter Memorial Fund  

A Computing Cluster Implementation Using SLURM

Joel Detweiler ('12); Matthew Bradley; Mentor: Everett Bull

Abstract:  As computers become more powerful, they are capable of solving more computationally expensive problems. Computer clusters are designed to solve many of these problems together and simultaneously. However, managing a cluster of computers and ensuring efficiency is a difficult task. My goal was to assemble a cluster of computers using an efficient resource management system that would also have an easy interface for students to submit jobs. After much exploration and testing, I set up a SLURM (Simple Linux Utility for Resource Management ) implementation on a cluster of 8 dual-core hosts. This optimized implementation allows for the submission of thousands of jobs which are processed based on available resources and their place in the queue.  Our SLURM cluster met our goals while also allowing for easy expansion in the future.
Funding Provided by: Pomona College SURP  

The Implementation of Object-Oriented Languages in Pedagogical Programming Environments

Chris Fincher ('12); Mentor: Kim Bruce

Abstract:  The Grace programming language project was started with the intention of making a new object-oriented language for teaching the practice of programming. In order to be successful, Grace must be approachable, and a significant factor in approachability is having a beginner-friendly integrated development environment (or IDE). We decided to investigate DrRacket as an IDE for the use of Grace, due to its numerous novice-friendly features. While DrRacket is built on functional, rather than object-oriented, languages, Java, an object-oriented language, was implemented in DrRacket as part of the ProfessorJ project. Unfortunately, the ProfessorJ code was in need of revision, and was not compatible with the newest version of DrRacket, but the codebase was brought back up to working specifications during the course of the project. Additionally, we prepared a Racket lexer and parser for a draft of the Grace specification that should allow the DrRacket engine to take in Grace code.
Funding Provided by: Paul K. Richter and Evalyn E. Cook Richter Memorial Fund   

Research at Pomona