N-Gram Generator

An n-gram is simply a sequence of n consecutive words, and the reason they matter is that meaning frequently lives in phrases rather than in single words. "Machine" and "learning" tell you very little separately; "machine learning" is a topic. Single-word frequency analysis systematically misses this, which is why a word count can suggest a page is about one thing while its bigrams reveal it is about another.

Bigrams and trigrams are the useful range in practice. Unigrams are just word frequency, and anything above four words appears too rarely in a single document to be informative unless you are analysing a large corpus or hunting for duplicated passages. Filtering out n-grams that consist only of stop words removes most of the noise, since "of the" and "in the" will otherwise dominate any list.

The most practical application is finding anchor text and internal linking opportunities: recurring bigrams across a set of pages are usually the concepts that deserve their own page and inbound links from the others. The second is style auditing — a phrase appearing eleven times across an article is a verbal tic the author cannot see, and n-grams surface it immediately where rereading does not.