High Performance PostgreSQL for Rails

Using Trigrams with FTS

In this section, you’ll work with trigrams, which are another way to search text. To do that, you’ll use the pg_trgm^[418] extension mentioned (and enabled) earlier.

If you’re not familiar with trigrams, let’s cover what they are.

Words are made up of sequences of letters. These sequences are called n-grams^[419] in the field of Natural Language Processing (NLP). N-grams are sequences without a specified length. An n-gram with a length of one is a unigram, two is a bigram, and three is (drumroll, please) a trigram.

When two words have many overlapping trigrams, they tend to be very similar. Consider the example of misspellings. A misspelled word will usually be very similar to the target word, sometimes off by a single ...