To Do

The people whose participation and advice has been solicited and/or volunteered for The Archive project are:

• Anwar Adi
• Dan Berger
• Dimitrios Gunopolus
• Victor Hill
• Eamonn Keogh
• Vinh Lam
• Mart Molle
• Walid Najjar
• Tom Payne
• Ravishankar
• Johannes Ruscheinski
• Wagner Truppel
• Titus Winters

Data caputure.  The following tasks urgently need to be accomplished for the capture portion of The Archive:

• Directories and TurnIn
  • Decide on layouts of central directories -- see ``our model''.
  • Construct skeleton directories per that layout.
  • Decide on authorization/permissions policies.
  • Modify TurnIn to use ACLs rather than special accounts such as prof, ta, and grad.
  • Design versioning crawlers and retrievers. (Can we somehow use the current versioning-based on-line backup system? There should be no need for data compression.)

• Spreadsheets
  • Decide on formats in detail (including normalization info).
  • Script to construct, initialize and update per-section score spreadsheets from corresponding rosters.
  • Instructions to TAs on how to maintain score spreadsheets.
  • Script that makes available to each student his/her own scores in some suitable format with appropriate contextual information such as min, max, and mean. (Not super urgent.)
  • Routine to populate a column of the spreadsheet with scores for the corresponding item.
  • An on-line test proctor/scorer, which must use the above routine. (Perhaps BlackBoard and/or John Gerdes' system can be used.)
  • Anwar's autograding stuff, which must also use the above routine.

• Question banking
  • Tools for test analysis:
    • LaTeX: parse exam class.
    • Word: study and possibly adapt Walid's exam template.
    • GUI test tool: Titus and Vinh.
  • format of item bank. (Wagner is working on this.)
  • format of test representation.
  • Tools for test synthesis. Need to be able to ask for all (or a random selection of) the items having attributes in certain ranges, which is an operation for which relational databases are especially well suited.

• Design and implement an extractor script that builds (say within a database) from the above information:
  • a sparse int array of when each student attempted each item that s/he attempted.
  • a sparse float array of how each student did on each item that s/he attempted.
  • a sparse float array of the relevance of each item to each desirable competence.

Data analysis.  The next phase is data analysis, which has a more liesurely schedule and for which the services of Eamonn, Dimitrios, Mart, and Ravi have been solicited. Initially we need alogorithms that use the three extracted arrays and say the methods discussed in [R29] and [R30] to build:

• an updated relevance array.
• a float array giving each student's expected level of each desirable competence.
• various per-item statistics that can be useful in selecting question when constructing tests.