This work proposed a new method to classify text documents
using fuzzy rules in order to consider the relevance
inherent imprecision of the documents in each topic. The
experiments showed that this is a promising approach to deal
with the problem of imprecision and dimensionality when
use rules to classify text documents. Moreover, considering
that the Wang&Mendell method is a grid-based method and
that one of its problems is that it usually produce a large
number of rules, for this method the number of rules was
very good, since the number of rules were about 5% of
the number of documents. However, investigations to be
continued in the future include the experiments with different
domains, preprocessing of the documents and analysis
about the reduction in the number of rules. Further, we also
can compare the results obtained by the method proposed
with results from the documents classification by a Support
Vector Machine method. Currently, the authors are working
in an extension of the method to do information retrieval
after the document organization using rules, in order to
do inference considering that the same document can fire
different rules but with different degrees.