An N-gram is an ordered collection of N elements of the same kind, usually presented in a large collection of many other similar N-grams. The individual elements are commonly natural language words, though N-grams have been applied to many other data types, such as numbers, letters, genetic proteins in DNA, etc. Statistical N-gram analysis is commonly performed as part of natural language processing, bioinformatics, and information theory.
An N-gram is an ordered collection of N elements of the same kind, usually presented in a large collection of many other similar N-grams. The individual elements are commonly natural language words, though N-grams have been applied to many other data types, such as numbers, letters, genetic proteins in DNA, etc. Statistical N-gram analysis is commonly performed as part of natural language processing, bioinformatics, and information theory.
N-grams may be derived for any positive integer N. 1-grams are called "unigrams," 2-grams are called "digrams," 3-grams are called "trigrams," and higher order N-grams are simply called by number, e.g. "4-grams". N-gram techniques may be applied to any kind of ordered data. Metadata such as end-of-sentence markers may or may not be included.
For example, using words as the elements and an N of 2, the English sentence "Three cows eat grass." could be broken into the 2-grams [{Three cows}, {cows eat}, {eat grass}, {grass #}], where # is a metadata marker denoting the end of the sentence.
As N-gram analysis embeds the data set into a vector space, it allows the application of many powerful statistical techniques to data for prediction, classification, and discernment of various properties.
More information:
