Some tips if you want to develop it for yourself:

You have to weigh the words by their source. The following is an example preference: Words from title: 100, words from meta keywords: 75, words from headlines: 50, words from the document body: 10.

Store the place of the word in the given document. E.g. the word is in the first 100 words. When the user is looking for two or more words, you can compare the hits, as if the found words are closer to each other as the other findings. Then you can compute a rank.