RRI step one: generate document vectors

For each document:

For each concept in document:

S(document)+ = E(concept) × lw × gw

where:

lw = log(1 + frequency of concept in document)

gw = entropy(concept)

$\text{entropy}=1+\Sigma \frac{p_{\mathit{ij}}{\mathit{log}}_2\left(p_{\mathit{ij}}\right)}{{\mathit{log}}_2\left(n\right)}$

$p_{\mathit{ij}}=\frac{{\mathit{cf}}_{\mathit{ij}}}{{\mathit{gf}}_i}$

cfij = frequency of concept i in document j

gfi = frequency of concept i in corpus

For each semantic document vector:

Normalize (majority rule)

RRI step two: generate semantic concept vectors

For each concept:

For each document concept occurs:

S(concept)+ = S(document)

For each semantic concept vector:

Normalize (majority rule)