2 TRAINING AND EXECUTIONDOCUMENT FEATURE SET (DF) IS A FEATURE SET E...

3.2 Training and Execution

Document Feature Set (DF) is a feature set ex-

The training phase estimates a probabilistic model

tracted only from a document. Using only DF corre-

from training data (x

⁽¹⁾

,y

⁽¹⁾

),...,(x

⁽ⁿ⁾

,y

⁽ⁿ⁾

) gener-

sponds to unbiased Term Extraction (TE).

ated from the CRL QA Data. The execution phase

For each word w

i

, the following features are ex-

evaluates the probability of y

⁰⁽ⁱ⁾

given inputx

⁰⁽ⁱ⁾

us-

tracted:

ing the the probabilistic model.

dw–k,. . .,dw+0,. . .,dw+k: k preceding and follow-

ing words of the word w

_i

, e.g., { dw–1: w

_i

−

1

,

Training Phase

dw+0:w

i

, dw+1:w

i+1

} if k = 1,