Clustering based approach extracting collocations
This addresses collocation extraction for natural language processing, but it appears incremental as it builds on classical measures and clustering methods.
The study tackled collocation extraction by using a clustering technique to separate bigrams into groups based on collocation probability, allowing exclusion of unlikely groups to reduce search space meaningfully.
The following study presents a collocation extraction approach based on clustering technique. This study uses a combination of several classical measures which cover all aspects of a given corpus then it suggests separating bigrams found in the corpus in several disjoint groups according to the probability of presence of collocations. This will allow excluding groups where the presence of collocations is very unlikely and thus reducing in a meaningful way the search space.