The number of pages required to build a contextually-relevant semantic index depends on the search term.
Some search terms yield large amounts of text-rich pages which create enough of a corpus for adequate semantic analysis in very few pages.
In other cases, where the search results contain a large number of text-light pages (as occurs with shopping sites, image-heavy pages or single-page applications) Wordmetrics will automatically do a "deeper dive" into search results until the size and richness of the corpus is sufficient to make the necessary statistical calculations.