The upper 30 ranked documents from the results of the
first retrieval are grouped according to classification
information and the term clusters that represent each
classification group are generated. The terms that represent
each classified group are extracted and then the term cluster
is generated. Next, the user to select term cluster that fulfills
user information needs with reference to the classification
information, ratio of the retrieved document (document
ratio), and the list of terms, such as the classification
information provided in Table 3. The term cluster provided
to a user shows top seven terms in the list of term cluster
according to the weight of tf*idf.