Where N is the number of tokens in the page.
This can be done simply by scanning the possible values for i and j and computing the objective function.
Note that this procedure will only work when the proportion of text tokens in the non-content section is lower than the proportion of tags,
which is not the case for the web page in Figure 3.17.
Pinto et al.
(2002) modified this approach to use a text window to search for low-slope sections of the document slope curve.