Collocati

Collocati, in linguistics, are the words that frequently co-occur with a given target word within a corpus. The term is the plural form of collocato in Italian and is often used interchangeably with the English term collocates. Identifying collocati helps reveal how words patternally combine in real usage and how meaning is built through fixed or conventionalized phrases.

Collocati are typically determined by examining a defined window of context around the target word, such as a few words to the left and right. Statistical measures are used to assess the strength of association between the target word and each surrounding word. Common metrics include mutual information, t-score, and log-likelihood, sometimes supplemented by more modern methods from distributional semantics. A higher measure indicates a stronger collocational relationship.

Applications of collocati analysis include lexicography, corpus linguistics, language teaching, and natural language processing. It aids dictionary makers in documenting common phrases, helps learners acquire idiomatic usage, and supports algorithms for parsing, machine translation, and sentiment analysis by prioritizing statistically salient word pairings.

Limitations and caveats exist. Collocati depend on the size, genre, and register of the underlying corpus, so results can vary across corpora. Polysemy and domain shifts may blur collocations, and not all adjacent words are true collocates. Consequently, researchers often use multiple measures and manual inspection to validate findings and consider varying contexts when interpreting collocational data.