Binary independence model

Creator

Created

2025 Jan 29 12:56

Editor

Edited

2025 Jan 29 12:58

Refs

IR as Classification (relevant or not) with

Assumes the relevance between documents and queries as binary variables (0 or 1)

P(D | R) = \prod_{i: d_i = 1} p_i \prod_{i: d_i = 0} (1 - p_i)

P(D | NR) = \prod_{i: d_i = 1} s_i \prod_{i: d_i = 0} (1 - s_i)

S(D) = \log \frac{P(D | R)}{P(D | NR)}

S(D) = \sum_{i: d_i = 1} \log \frac{p_i}{s_i} + \sum_{i: d_i = 0} \log \frac{1 - p_i}{1 - s_i}

////////////