digram

A digram is a unit consisting of two adjacent elements within a sequence, such as two consecutive letters, digits, or other symbols. In linguistic analysis and information theory, digrams are studied as two-letter sequences, revealing patterns in a language or corpus. The term is sometimes used interchangeably with bigram, but authors may reserve digram for letter pairs or other small-scale units.

In text analysis, digrams are the simplest form of an n-gram and are used to model structure and predictability in text. To construct a digram model, each pair of consecutive characters or tokens is counted, and probabilities are assigned to what often follows a given symbol. This approach helps measure similarity between texts, detect language, or inform compression and cryptanalysis.

Example: the word "example" yields the digrams "ex", "xa", "am", "mp", "pl", "le". In longer corpora, common English digrams include "th", "he", "in", and "er".

In cryptography, digrams have historical relevance in systems that operate on pairs of letters, such as digraphic and Playfair-type schemes, where the security depends on pairwise letter frequencies. In modern NLP, bigram models often refer to pairs of tokens (words or characters) and are a standard building block for language modeling.

Because the term digram is sometimes used variably, readers should check context and field. See also bigram, trigram, n-gram.