How do you train a search engine on books? How do you generate relevance metrics (like # of links on the web)?
Ie when you put in a word 'Mroscow' it sees that this term has 500 results while a similar term 'Moscow' has 100 million results. Therefore it assumes that you simply misspelled the word and offers the alternative as suggestion.
It does not know what the actual underlying word means.
PageRank is one of the 200 metrics Google uses nowadays. It's weight is not as great before.