August 17, 2005
Generic Soft Pattern Models for Definitional QA
13/28
Bigram Soft Pattern Model
Bigram prob
Slot-aware unigram prob
 P ( Ins ) = P(“known”|S-2) + P(“as”|S-1) + P(“,”|S1) + P(“DT$”|S2) + P(“known as”) + P(“, DT$”)
•To estimate the interpolation mixture weight λ
–Expectation Maximization (EM) algorithm
•Count words and general tags separately
–Avoid overwhelming frequency count of general tags
Here we consider the pattern matching problem as a token sequence generation problem. So, we take the token sequence t1 till tL from the test instance and calculate its probability according to the training data. In a typical bigram model, this generation prob is multiplication of bigram probs. Here, we use linear interpolation to smooth the bigram probs and introduce two terms….