DATA-DEPENDENT CLUSTERING OF GEOSPATIAL WORDS

摘要

Embodiments include method, systems and computer program products for data-dependent clustering of geospatial words. In some embodiments, a matrix of words and locations may be obtained. A plurality of locality-sensitive hash (LSH) functions may be generated. A plurality of sparse row vectors may be generated. A plurality of sparse LSH functions may be generated based on the plurality of LSH functions. A clustering bucket may be determined from a plurality of clustering buckets for each of the plurality of sparse row vectors using the sparse LSH functions.