Calculating Similarity
¡
Euclidean Distance - bad
l
M(Q,D
d
) = sqrt (
Σ
|w
q,t
– w
d,t
|
2
)
l
Dissimilarity Measure; use reciprocal
l
Has problem with long documents,
why
?
¡
Actually don’t care about vector
length, just their direction
l
Want to measure difference in direction