Figure 2.

GO node Word Expansion via proximity measure. (1) For each document, a Boolean matrix of word occurrence in paragraphs (R) is created. (2) Co-occurrence proximity network WPP is computed. (3) Words in GO node label (WGO) are expanded (WGOProx) using WPP. (4) Intersection of vector of expanded GO node words (WGOProx) with word vectors for each paragraph in the document (columns of R) : paragraph with largest intersection is returned.

Verspoor et al. BMC Bioinformatics 2005 6(Suppl 1):S20   doi:10.1186/1471-2105-6-S1-S20