Complementarity and Similarity: Relationships Between Text-Mined Terms and Social Tags for Image Description

View/ Open
Date
2010-07-19Author
Klavans, Judith L.
Cho, Hyoungtae
LaPlante, Rebecca
Metadata
Show full item recordAbstract
In this paper, we present our results on comparing the language of
social tags with text-mined terms for images. We have developed a novel
modification of the standard term frequency/inverse document frequency
metric (tf*idf) (Salton & Buckley 1988) over tags and terms to identify
and filter terms which discriminate images for searchers. Since tags
serve as additional input, we refer to this modification as the T-tf*idf
Measure, i.e. Tags-term frequency as an inverse of document frequency,
where "document" in this case refers to the either the tag or term
dataset. We present the results of several variations on this measure,
and demonstrate the impact on output. We discuss evaluation of our
results on the ability of the metric to reflect human judgments through
experiments which illustrate the value of the approach.