Title: Latent Semantic Indexing
Performer(s):
Walter Kintsch
Institute of Cognitive Science
University of Colorado
Campus Box 344
Boulder, CO 80309-344
Tom Landauer
Institute of Cognitive Science
Unviersity of Colorado, Boulder
|
Cluster: SNAIR
Contact Information:
Phone: 303-492-8663
Fax: 303-492-7177
email: wkintsch@clipr.colorado.edu
Phone: 303-492-2875
email: landauer@psych.colorado.edu
|
1. Instructional Focus:
Content areas/topics: None specified.
Process skills: None specified.
2. Target Population: Best for grades 4-12 since available text is most appropriate to these grade levels, however generally applicable to any reading level.
3. Summary Description: The project will develop educational applications of Latent Semantic Indexing (LSA), a new machine learning method for characterizing, comparing, matching, and retrieving textual materials.
The performers will research and develop methods for automatically selecting instructional text that matches both the detailed topic and the optimum level of discourse sophistication for individual students, and develop a prototype of a network-based search and delivery system. The methods to be investigated rely primarily on Latent Semantic Analysis, a high-dimensional model that extracts and represents the meaning of words by analysis of large bodies of text.
The product will match students to instructional text (and other materials) on their current topic of interest and at their current level of sophistication. This should help to make the use of large electronic tect resources (e-libraries) more valuable.
4. Training and Staff Development:
- Teacher prerequisite skills/knowledge needed: None specified.
- Student prerequisite skills needed: None specified.
- Training needed/provided: None specified.
- Technical support needed/provided: None specified.
5. Technological/Resources Needed: Sun/Solaris servers; almost-anything clients; PC/Windows probably preferred, but delivery to X/Mosaic/WWW front ends is anticipated.
Most applications can probably be configured as low-end clients over Internet to our server resources where maximum possible processor and memory resources are desirable. (A SparcStation model 72 with 512M main memory is needed on our end at a minimum.
6. Intended Outcomes:
Students: None specified.
Teachers: None specified.
7. Instructional Time Required: Not applicable.
8. Role of the Pilot Teacher(s):
9. Example(s) of the Use of this Product (Scenario):
A teacher has acquired a very large amount of text information from the Internet. Some, but not all, of it could be used for a curriculum unit he is working on. He uses the tools developed in this project to search this large collection of information for items usable in his curriculum unit.
You want a quick, automatic way to grade short essay exams? We will be able to approximately help.
You want to evaluate whether instructional text is coherent and understandable? We will be able to show you how.
Students guided to information which content relevant based upon student model.
Teacher uses grading tool incorporating automated scoring of student papers.
