This thesis proposes a novel model for automatically generate topic map for a document corpus with no supervision. We extend a previous approach to discovery of lexical relations from text data to construct a hierarchy of ...