Abstract:
Semantic-based information retrieval in the digital repositories is becoming
an important mechanism to facilitate end users with an ease of exploring
intensive volume of information. The traditional keyword-based approach
of retrieving information does not classify and conceptualize the context for
searching digital data owing to the fact that these approaches are based on
literal matching. Consequently, the end users put signi cant e orts to arrive
at the required information even if it exists in the search space. The main
focus of this thesis is to exploit the semantics of taxonomy in order to improve
the results of subject searching in institutional repositories.
Our proposed system uses taxonomy for subject-based searching in dig-
ital repositories. In proposed system the documents need to be annotated
either manually or automatically on subjects of the taxonomy. The annota-
tions play key role in searching documents on the basis of taxonomy subjects.
Ontology provides the semantic understanding of the relationships that ex-
ist among di erent objects of a domain. We deployed the ACM Computing
Classi cation System (CCS) taxonomy in ontological format in order to ex-
ploit the semantic relationships that exist among the taxonomical subjects.
In the proposed searching system, we consider the family of the selected sub-
ject instead of only the exact subject for searching. By family of a subject, we
mean immediate broader and narrower, and related subjects. Moreover, we
apply ranking algorithm which assigns di erent weights for exact, broader,
narrower and related matches of the subject. The retrieved documents are
ranked according to the calculated scores. We compared our proposed sys-
tem with Controlled Vocabulary add-on developed in DSpace - institutional
digital repository, for subject searching of documents. We evaluated the sys-
tem on basis of precision, recall and F-measure and found the results of our
proposed system very promising.
