N-gram

A short sequence of word types

An n-gram is a short sequence of word types. n is the number of words in the sequence, for example, a 2-gram has two word types. N-grams have many applications in machine translation:

language models, such as an n-gram maximum likelihood estimate
translation models for statistical machine translation
evaluation metrics, such as BLEU
language identification

Number of words	Common name	Sequence notation	N-gram language model notation
1	Unigram	$(w)$	$P(w)$
2	Bigram	$(w_1, w_2)$	$P(w \vert w_{-1})$
3	Trigram	$(w_1, w_2, w_3)$	$P(w \vert w_{-1}, w_{-2})$

Example

String in English: "The car has two doors."

Tokens: "The", "car", "has", "two", "doors", "."

Unigrams: "The", "car", "has", "two", "doors", "."

Bigrams: ("The", "car"), ("car", "has"), ("has", "two"), ("two", "doors"), ("doors", ".")