A trigram is a contiguous sequence of three adjacent words found within a text string. As a type of N-gram, it is a key component in natural language processing models used to analyze word patterns and sequences in search queries.
Trigrams are utilized in ML-assisted search intent classification systems. These systems parse the user query using techniques like phrase detection and then leverage models that specifically handle bigrams and trigrams to perform semantic similarity recognition. This helps the system determine the user’s intent by understanding common phrases and sequential terms that suggest a specific goal, such as looking for a “short fact” or “instruction”.