|
For an incoming
table, we construct it’s own word set. We then get the effective word set by
finding the intersection of the generic word set and the new individual word
set, and denote it as .. Where m_1,
etc are indexes back to the generic work set. We then compute three vectors.
Vector Gs represents the genuine table class and is composed with the
corresponding genuine table weights divided by a normalization factor. Vector
Ns …. And vector It represents the incoming table and …Finally the word group
feature is defined as the ratio of the dot product …. Basically, it measure
whether the vector representing the incoming table correlates better with the
genuine table vector or the non-genuine table vector.
|
