Corpora: automatic thesaurus generation from large corpora

[Adam Przepiorkowski]

Dear List Members,

Does anbyody know of attempts to automatically build hierarchical thesauri
from larga corpora using unsupervised or only minimally supervised methods?
Most of the work around concentrates on finding related terms (e.g., using
various clustering algorithms), but apparently there hasn't been much work
on building hierarchical thesauri, with (at least) the RT, BT and NT
relations and with all nodes in the hierarchy corresponding to meaningful
descriptors.  Any pointers?

Thanks in advance,

       Adam P.