Adaptive Information Filtering - Text Mining: Classification, Clustering, and Applications - page 215

Symbol Description

x DCG ( n )the Discounted Cumu-

lated Gain of a ranked list
of n passages
DCU ( n )the Discounted Cumu-
lated Utility of a ranked
list of n passages
NDCU the Normalized DCU
PNDCU the Penalized NDCU
G ( d i ,q )

a TF-IDF vector, repre-
sentating a passage

y the class label of a pas-

sage, indicating its rele-
vance (yes or no) to a
query

w ∗ a vector of regression coef-

ficients, serving as a query
profile or “class model”

the gain for reading doc-
ument d i with respect to
query q

f RL the solution of regularized

logistic regression, the
mapping function from
any passage to the
estimated conditional
probability for the passage
to be relevant to a query

U ( p i ,q ) the utility for reading pas-

sage p i with respect to
query q

C ( p i ) the set of nuggets con-

tained in passage p i

h i

a TF-IDF vector, repre-
senting a historical pas-
sage

β the dampening factor that

penalizes re-occurrences of
a nugget in ranked pas-
sages

H ( t ) the user history at time t ,

defined over a sequence of
historical passages

λ the weight balancing

the NDCU term with

f ND ( x )

the novelty scoring func-
tion applied to an input
passage

the penalty on ranked-

list length in utility

assessment

f AR ( x ) the anti-redundancy scor-

ing function