| Summary: | 碩士 === 中原大學 === 資訊工程研究所 === 82 === In this thesis, a new approach, called mixed approach, is
proposed for reusing components from software library. The
mixed approach is essentially a hybrid of faceted index and
free-text analysis method. The motivation of this research is
an attempt to use both faceted index and automated retrieval
approach in reusing software components from a C++ class
library. A CIG, Conceptual Information Group, is a group of
significant words that represent fuzzy concepts. An extended
faceted classification allows the reuser to classify the class
components with four CIGs. The reuser can present his required
concepts with CIGs while retrieving. The evaluation of an
information system includes parsing the target class source
code and counting their frequency with the free-text analysis
approach, which depends on these significant words described in
CIGs and their frequencies. Then the system transforms the
frequencies into standardized scores. An FCIG, Frequency of
CIG, is the group of standardized scores of CIG. Candidate
classes are selected from the library based on the degree of
similarity between their CIGs and the required descriptions.
Similarity is quantized by a standardized score of matched
significant words. We implement a prototype browser to work
with an assembled C++ class library, which is based on the
mixed approach. Finally, we performed two experiments on
retrieval to support our idea. A precision criterion is
proposed. Also, some interested experimental results are
discussed.
|
|---|
