Pullman's Muse

Published: 06-07 2005
Title: Music and content analysis
Topic: Robotic grading

There's an article at Yahoo! News about a company called Savage Beast Technologies that " combine[s] computer analysis with the art of listening to come up with new music suggestions for consumers based on what they already like."

Their job is to discern and define attributes in tunes by artists as diverse as teen diva Hilary Duff and jazz legend Miles Davis. The listeners classify hundreds of characteristics about each song, including beat, melody, lyrics, tonal palette and dynamics, then plug the data into a music recommendation engine — software designed to find songs that share similar traits. "It's about understanding someone's music taste," said Savage Beast founder Tim Westergren. "Why does somebody like a piece of music?"

The description of how it works (as well as the quotations from some in the music industry expressing skepticism about its accuracy) remind me of the content analysis programs I've been looking at lately. Is it really possible to code music, to generate the meta discription that would categorize music, all music, even before a piece is written, so as to match a piece to its inevitable audience? (link)

comment | bookmark | backlinks | edit | delete

Published: 06-01 2005
Title: Latent Semantic Indexing software
Topic: Robotic grading

Telecordia Technologies offers a software program that uses singular value decomposition to determine the relevance of multiple texts based not on key word comparisons but concepts. As they point out (link), if a searcher looks for "laptop" he won't find articles about "portables" or "palm pilots" nor any of the other synonymous words. Their search tool creates concepts of words (synonyms?) and searches based on those concepts rather than on any given word per se.

comment | bookmark | backlinks | edit | delete

Published: 06-01 2005
Title: LSA article
Topic: Robotic grading

Peter W. Foltz. New Mexico State University (1996) "Latent Semantic Analysis for text-based research". Behavior Research Methods, Instruments and Computers. 28(2), 197-202. (link)

Abstract

Latent Semantic Analysis (LSA) is a statistical model of word usage that permits comparisons of semantic similarity between pieces of textual information. This paper summarizes three experiments that illustrate how LSA may be used in text-based research. Two experiments describe methods for analyzing a subject's essay for determining from what text a subject learned the information and for grading the quality of information cited in the essay. The third experiment describes using LSA to measure the coherence and comprehensibility of texts.

Salient conclusion:

The grades assigned by the second LSA measure, (overlap with expert model), correlated well with three of the four graders and the correlations were stronger than the first measure. So, the quality of an essay can be characterized as a match between the expert's model of the domain and what was written in the essay. Indeed, the graders' correlations with LSA expert model are well within the range of the correlations between the graders.

comment | bookmark | backlinks | edit | delete

Published: 05-11 2005
Title: A CNN article on Missouri-Columbia Soc. Prof
Topic: Robotic grading

"COLUMBIA, Missouri (AP) -- Student essays always seem to be riddled with the same sorts of flaws. So sociology professor Ed Brent decided to hand the work over to a computer."

(link)

comment | bookmark | backlinks | edit | delete

Published: 04-15 2005
Title: Robotic writing
Topic: Robotic grading

CNN, via metafilter, has a nice follow up to my last post.

CAMBRIDGE, Massachusetts (Reuters) -- In a victory for pranksters at the Massachusetts Institute of Technology, a bunch of computer-generated gibberish masquerading as an academic paper has been accepted at a scientific conference.

(link)

comment | bookmark | backlinks | edit | delete

Published: 04-12 2005
Title: The other side of the equation
Topic: Robotic grading

I sometimes joke that once we perfect the electronic grader, all we will need is a robotic writer and we will have finally eliminated human beings from the practice of first year composition. Well, here's a link to beta 1. (link)

comment | bookmark | backlinks | edit | delete