LSA: First dimension and dimensional weighting

X. Hu, D Franceschetti, P. Penumatsa, AC Graesser, M Louwerse, D. S. McNamara, Tutoring Research Group

Research output: Chapter in Book/Report/Conference proceedingChapterScientificpeer-review


We report two discoveries concerning Latent Semantic Analysis (LSA). First, we observed the special properties of the first dimension of the LSA space. Second, we observed that dimensional weighting plays an important role in LSA analysis. Based on the first discovery, we examined the cosine matches without the first dimension. Based on the second discovery, we explored different dimensional weighting schemes. Based on these observations, we recommend a new algorithm for LSA cosine computation such that LSA becomes more sensitive to relevant similarities and differences.
Original languageEnglish
Title of host publicationProceedings of the 25h Annual Conference of the Cognitive Science Society
EditorsRichard Alterman, David Kirsh
PublisherCognitive Science Society
Publication statusPublished - 2003
Externally publishedYes
Event25th Annual Conference of the Cognitive Science Society - Boston, United States
Duration: 31 Jul 20032 Aug 2003
Conference number: 25


Conference25th Annual Conference of the Cognitive Science Society
Abbreviated titleCogSci 2003
Country/TerritoryUnited States
Internet address


Dive into the research topics of 'LSA: First dimension and dimensional weighting'. Together they form a unique fingerprint.

Cite this