CS6604 Projects: Suggestions

There are many possible projects suitable for CS6604. Please read over the Overview of Related Projects. Then select from the Suggestion List below, or suggest an adaptation thereto, or propose a project of your own creation.

To learn about personality issues that may help your work more effectively with other groups members, see a variant of the Myers Briggs personality test that you can take in about 5 minutes.

Overview of Related Projects

• CS Courseware DL
  NSF has supported our efforts to improve CS education by making our CS digital library accessible to students, and by funding faculty to make learning more interactive (adding online interactive exercises or making tools or packages available).
• Envision Interface
  As part of our work on digital libraries, the Envision interface was developed. It includes a query window, a results list window, and a results visualization window. Some other features have been designed but not implemented, such as browsing or zoom in/out.
• IBM DL
  IBM is working with Virginia Tech on Digital Library research and development. This involves use of search engines, rights management software, DBMS, and image handling software. Extensions desired relate to making it compatible with NCSTRL, NDLTD and with the Envision Interface.
• MARIAN
  MARIAN is a library catalog search system developed at Virginia Tech. It now runs with the same content as VTLS, and serves as a vehicle for research as well as providing an alternative to VTLS.
• NCSTRL
  The Networked CS Technical Report Library runs as a distributed digital library among CS departments. Further research is needed to improve its performance and reliability.
• NDLTD
  Virginia Tech is funded by Southeastern Universities Research Association and US Dept. of Education to build a National Digital Library of Theses and Dissertations. This involves developing a distributed digital library of great size, training large numbers of students about digital libraries and electronic publishing, and developing tools and procedures so students can easily add to the archive.
• VSE
  Simulations can help with understanding phenomena, especially when they run interactively and give visual feedback. Dr. Balci's Visual Simulation Environment (VSE), can be directly applied, and he will aid those interested in working with that tool. It should be of help in modeling and simulating digital library systems.
• ZPRISE
  The Natural Language Processing and Information Retrieval Group at NIST makes available a public domain package of source code and documentation (ZPRISE), which includes a Z39.50 UNIX client/server pair. It runs on Solaris. See online documentation. Other Z39.50 clients and servers exist, and might be used instead.

Suggestion List

• Title: EI Project Reporting
  • Number of people: 2
  • Goal: Article summarizing findings of EI project
  • Desired background: statistical reporting, educational evaluation
  • Contact: instructor
  • Description: The CS Courseware DL (see above) has had many successes which should be summarized in an invited journal article due January 1998. This can be done by contacting the 5 principal investigators (Barnette, Fox, Hartson, Lee, and Shaffer), getting data and results from them, and summarizing the activities and and findings.
• Title: CS Courseware DL - Log Analysis
  • Number of people: 2-4
  • Goal: Tools to capture and characterize users' sessions and accesses to courses.
  • Required background: PERL, UNIX, probability and statistics
  • Contact: instructor, Ghaleb Abdulla (abdulla@csgrad.cs.vt.edu)
  • Description: We have extensive logs of students accessing the CS courseware online for over 2 years. These can be analyzed with various routines already developed, as well as statistical analysis packages. Preliminary work in this regard has been undertaken by Binzhang Liu and Ghaleb Abdulla.

    We wish to prove if and how a digital library of this type can aid learning. We also wish to: discover how courses can be improved if digital library technology is available, prepare tools to help in that process, and generate recommendations that will yield the greatest benefit. We need to develop tools to extract users' sessions and characterize them.

    We will try to answer the following questions: Can we identify a user session and how? If yes, can we characterize it? What is the effect of a single session on the network and server? Do we have different types of sessions, or are they all the same? If they are different, what types of sessions do we have, and what do we need to support every type of session? How can user sessions be related to actual user behavior to determine what is viewed, what is studied in-depth, what confusions arise, what WWW design problems cause users difficulties, and what can be concluded from logs without having to synchronize them with user behavior? How do sessions combine so users can serve their information needs over longer periods of times, especially in distance education situations?

• Title: IR Query Log Analysis
  • Number of people: 2-3
  • Goal: Characterization of IR queries, building of large test collections, construction of models of query submission.
  • Required background: probability and statistics
  • Contact: instructor, Ghaleb Abdulla (abdulla@csgrad.cs.vt.edu)
  • Description: We have a large number of queries available from records of IR systems. One is the full log of MARIAN queries. Another is the log of NCSTRL queries. The third can be obtained from InfoSeek, one of the large companies providing support for searching the WWW. We would like to characterize these query collections, compare them, build models of them, and when possible, develop synthetic collections. This work can build on earlier work by G. Abdulla and B. Liu which identified sessions and: Browse, Search, Next-screen
• Title: Envision Interface
  • Number of people: 2-6
  • Goal: Design and if possible (partial) implementation of enhancements to the current Envision interface.
  • Required background: C, Java, AWT
  • Contact: Robert France, france@vt.edu, and Lucy Nowell, nowell@vtcc1.cc.vt.edu
  • Description:
    • The current Envision interface (see textbook Figure 7.6 - a color plate) should be redone in Java.
    • The Envision interface could be applied to data of interest to the National Library of Medicine.
    • The Envision interface could be adapted by connecting it say with NIST's ZPRISE system, to work with any Z39.50 accessible collection.
    • The Envision interface could be connected with a backend system to extract information from databases on the WWW. One approach to this is to use CPL from ISS at NUS in Singapore.
    • A detailed design of browsing for Envision could be completed, and implementation begun.
    • A detailed design of zooming (in, out) for Envision could be completed, and implementation begun.
• Title: IBM DL Enhancement for NCSTRL, NDLTD
  • Number of people: 2-3
  • Goal: IBM DL capable of fitting in with NCSTRL and NDLTD.
  • Required background: C++
  • Contact: instructor, Paul Mather (paul@csgrad.cs.vt.edu)
  • Description: NCSTRL follows an open architecture, with particular protocols used to connect the user interface, index server, and repository. Right now the only software usable with it is the Dienst package from Cornell. Using the API provided, the IBM DL system could be extended/adapted so that a site could use the IBM DL instead of the Dienst software.
• Title: NDLTD - IBM Digital Library Setup
  • Number of people: 2-5
  • Goal: The NDLTD must operate as a distributed digital library, with roles for Virginia Tech, UMI, and various universities. Use of IBM Digital Library software is essential so local knowledge of it can spread.
  • Required background: Java, C
  • Contact: instructor, Paul Mather (paul@csgrad.cs.vt.edu)
  • Description: As the coordinating institution for the NDLTD, Virginia Tech must set up the central digital library and prepare training materials for other universities. First, we must study, install, and adapt IBM Digital Library to the needs of NDLTD. Second, we must load data so that a full parallel / alternate service runs in addition to the regular library based system. Finally, we must document this so other universities can deploy our scheme and it will operate in distributed fashion.
• Title: NDLTD - Filtering
  • Number of people: 2
  • Goal: Design of a filtering mechanism for NDLTD.
  • Required background: systems analysis
  • Contact: instructor, Paul Mather (paul@csgrad.cs.vt.edu)
  • Description: Users of the NDLTD will want to have relevant new works brought to their attention, based on a stored profile that describes their interest. A design is needed for such filtering, at the level of institution as well as individual. The architecture of this service, what technology is used, and details of the user interface are needed.
• Title: Scaling DLs
  • Number of people: 1-5
  • Goal: An analytical model or software simulation of architectures that will enable DLs to scale to large numbers of users and large content objects.
  • Required background: probability and statistics
  • Contact: instructor, Ghaleb Abdulla (abdulla@csgrad.cs.vt.edu), Rob Akscyn (rma@ks.com)
  • Description: What will happen to digital libraries when serving millions of users with millions of gigabytes worth of information? The instructor has a preliminary formulation of a massively parallel solution, which needs to be completed and tested to see how various architectures will affect response time. Data from services such as InfoSeek can be obtained to allow some trace-driven simulations to help in this process. After characterizing users, sessions, clients, servers, and classes of servers and clients, we can use this data to build a predictive model. The model can be applied on various levels: disk level, server level, group of clients level, enterprise level, etc. If the model is accurate, then the invariants and characteristics identified can be used to test and build models for other Web communities.
  • Special Note: The "PetaPlex" digital library architecture of Knowledge Systems Incorporated may be studied with the assistance of Robert Akscyn, President of KS.
• Title: NDLTD - LaTeX and SGML
  • Number of people: 2
  • Goal: Students should be able to author theses and dissertations in LaTex and have them converted correctly to SGML, and SGML-coded theses should be convertible to LaTeX for formatting and printing.
  • Required background: LaTeX, some knowledge of SGML (or at least HTML)
  • Contact: instructor, Neill Kipp (nkipp@vt.edu)
  • Description: Based on the Document Type Definition for SGML, macros should be chosen so that an author working in LaTeX can use them to mark up a document. A converter must be developed to translate from LaTeX to SGML. Another converter must be developed to translate from SGML to LaTeX. (Note: work on this began in 1996, so that work needs to be extended to complete the task. Usability testing of the solution is required.)
• Title: Flash Crowd and Proxy servers
  • Number of people: 3
  • Contact: instructor, Ghaleb Abdulla (abdulla@csgrad.cs.vt.edu)
  • Description
• Title: 4S and DL Objects
  • Number of people: 2-3
  • Contact: instructor
  • Description
• Title: DL Implementation Group (Interface)
  • Number of people: 15
  • Contact: instructor, Neill Kipp (nkipp@vt.edu)
  • Description
• Title: NCSTRL Simulation
  • Contact: instructor, Ghaleb Abdulla (abdulla@csgrad.cs.vt.edu)
  • Description
• Title: Small Projects
  • Contact: instructor, Paul Mather (paul@csgrad.cs.vt.edu)
  • Description
• Title: Federated Search
  • Contact: instructor, ames Powell (jpowell@VT.EDU)
  • Description
• Background
  • Draft DLI2 proposal outline
• Outline for 8/28/97 class about projects
  • Outline for 8/28 class