Alexander Faaborg: Research and Projects at Cornell

 




Research > Research and Projects at Cornell
   

Computer Science Projects:

  Semantic Browsing
COLLS 499: Honors Thesis Research, Summer 2002, Fall 2002, and Spring 2003

Faculty Advisor: Carl Lagoze, Cornell University Department of Computer Science

This project provides a framework for annotating and reorganizing existing files, pages, and sites on the Web that is similar to Vannevar Bush's original concepts of trail blazing and associative indexing.  I have created several software applications that allow users to both author and use Semantic Web metadata. To create and use a layer of semantic content on top of the existing Web, I have (1) implemented a user interface that expedites the task of attributing metadata to resources on the Web, and (2) augmented a Web browser to leverage this semantic metadata to provide relevant information and tasks to the user.

ECDL 2003 Article
ECDL 2003 Presentation

Screen Shot: The Web Task Pane
Screen Shot: The Web Annotation Pane
Screen Shot: Site Annotator

Movie: Sequence Navigation
Movie: Downloading a Set of Images
Movie: Navigation on Semantic Links
Movie: Creating an RDF Triple
Movie: Creating an RDF Bag

  A USENET Style Interface for Peer-to-Peer Networks
CS 490: Independent Research, Spring 2002

Faculty Advisor: Carl Lagoze, Cornell University Department of Computer Science

LimeWire, a popular Gnutella client written in Java, has pioneered the use of rich XML queries to search the metadata of files residing on the Gnutella network. The Gnutella network is strongly search-centric.  There is currently no way to browse through information. Based on LimeWire's open source code, this project used LimeWire's XML query engine to create a USENET style interface on top of the Gnutella network.

Screen Shot
Paper
Presentation
  Link
CS 501: Software Engineering, Spring 2003

Faculty Advisor: William Arms, Cornell University Department of Computer Science

Project Team: Vincent Beurrier, Albert Eskenazi, Alex Faaborg, Peter Flynn, Riaz Jahangir, Michael Sorokorensky, Jeff Yuen

As our project for CS 501 Software Engineering, my team created an application to help intelligence analysts and law enforcement officers with the task of Link Analysis. Our application was designed to graphically represent associations between people, telephone records, financial transaction records, and other sources of information. The application allows analysts to import subpoenaed database records (often coming from a telephone company or financial institution), and provides tools for analysts to create graphical visualizations from scratch. Data can be exported as an XML file, making the application compatible with other Link Analysis tools currently on the market.

My main role on the project team was based on Microsoftís position of Program Manager. I designed the applicationís user interface after conducting interviews with analysts, and understanding how they did their job. My tasks included creating feature specifications based on user requirements, prioritizing features for our teamís developers, and tracking the teamís progress to ensure that we reached all of our milestones and delivered the application to our client on schedule.

Screen Shot
  Search Engine
CS 430: Information Discovery, Fall 2002

Faculty Advisor: William Arms, Cornell University Department of Computer Science

As the final assignment for CS 430, I created a search engine for the Web.  The search engineís crawler uses multiple threads to increase the speed at which it can index sections of the Web, and respects robots.txt files.  Parsed information is stored in an inverted index, and term frequencies are calculated using a standard tf.idf weighting scheme.  The search and discover service allows for Boolean and field-based searches, along with the ability to browse forward and backward links.  A
user interface (a parody of everyone's favorite search engine) was added for extra credit.

Screen Shot
Documentation

  Leveraging Metadata for Natural Language Processing
CS 473: Practicum in Artificial Intelligence, Fall 2001

Faculty Advisor: Bart Selman, Cornell University Department of Computer Science

The single fastest way to locate information online, or in any large body of documents, is with a text search. However, a pure text search is lacking in many regards. Often, documents are able to discuss topics while never directly stating them, or they will use slightly different terminology. A pure text search will scan documents for the occurrence of words, but it will follow no particular logic or reason in the results it returns. Recently XML and RDF have emerged to bring a semantic quality to information on the Web. While any human can look at a web page and immediately understand its semantics, XML and RDF are powerful because they provide semantic information that is understandable to machines. This project uses XML metadata to improve searching accuracy in the form of an interactive chatbot that is both significantly more intelligent than a pure text search, and provides a more natural user experience.

Screen Shot
Cornell BOOM 2002 Poster
Paper
   

 

Human Computer Interaction Projects:

  Using Neural Networks to Create an Adaptive Character Recognition System
COGST 416: Modeling Perception and Cognition, Spring 2002

Faculty Advisor: Michael Spivey, Cornell University Department of Psychology

A back-propagation neural network with one hidden layer was used to create an adaptive character recognition system. The system was trained and evaluated with printed text, as well as several different forms of handwriting provided by both male and female participants. Experiments tested (1) the effect of set size on recognition accuracy with printed text, and (2) the effect of handwriting style on recognition accuracy. Results showed reduced accuracy in recognizing printed text when differentiating between more than 12 characters. The handwriting style of the subjects had varying and drastic effects on recognition accuracy which illuminated some of the problems with the system's character encoding.

Screen Shot
Paper
  SamePage - Research In Real Time Collaborative Applications
COMM 499: Independent Research, Spring 2001

Faculty Advisor: Geri Gay, Cornell University Human Computer Interaction Group

Utilizing a network of client and server side applications communicating with http tunneling, samePage is able to facilitate real time push content and collaborative applications for use in an educational environment.

Paper
  Leveraging Metadata to Improve Information Retrieval in Directory Interfaces
COMM 499: Independent Research, , Fall 2001

Faculty Advisor: Geri Gay, Cornell University Human Computer Interaction Group

This study describes how metadata can be used to organize documents into hierarchical structures that filter against each other. It then discusses several experiments that were conducted to test the underlying usability concerns of this type of organizational system.

Paper
   

 

Computer Graphics and Vision Projects:

  A Psychophysical Study of BRDF Based Lighting
COGST 201: Cognitive Science in Context, Spring 2001

Faculty Advisor: David Field, Cornell University Department of Psychology

Subjects were presented with three-dimensional objects rendered with BRDF-based lighting. The subjects were tested to see if (1) the object they were looking at affected their ability to evaluate an object's material and (2) if they were able to tell different materials apart based only on light reflectance. Results showed that even though most subjects felt like they were guessing, subjects were able to correctly evaluate the objectís material with about 70% accuracy. Readers should note that this study used a reasonably small sample size that was not very unique demographically.

Test Chalice
Paper
  Visual Acuity in Foveal and Peripheral Vision
PSYCH 342: Human Perception: Applications to Computer Graphics, Fall 2001

Faculty Advisor: David Field, Cornell University Department of Psychology

We rarely notice that our peripheral vision is low resolution because we are constantly focusing on what we want to look at with our high resolution fovea. The muscles surrounding the eye allow us to quickly scan our 2 degrees of foveal vision to whatever area we need to view. Because of this, we are often completely unaware of just how low resolution our peripheral vision is. This project provides several interactive examples that demonstrate the contrast in resolution between foveal and peripheral vision.

Web Site
   

Research at MIT