Document And Query Expansion Method With Dirichlet Smoothing Model For Retrieval Of Metadata Content In Digital Resource Objects
Loading...
Date
2020-03
Authors
Alma’aitah, Wafa’ Za’al Mohammad
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Digital resource objects (DRO) refer to information that are structured which
elaborate, describe, and ease retrieval, usage and management of information
resources. Lately, the need for accessing the content of DROs has been addressed
differently by data retrieval (DR) and information retrieval (IR) research communities.
DR is found to be inadequate in providing enriched metadata content and may fail to
enhance the retrieval performance. In this thesis, an IR framework is proposed which
consists of three main stages: enhanced document expansion (EDE) method, adaptive
structured Dirichlet smoothing (ASDS) model, and semantic query expansion (SQE)
method. The first stage involves proposing an EDE method in which a new procedure
is introduced to increase each metadata unit content according to some specific steps
by adding new information which is more relevant and closer to each metadata unit in
each document while the second stage involves proposing an ASDS model that has
two scenarios to improve the Dirichlet smoothing model. The first scenario is to
enhance the model by taking into account of the document structure as in the proposed
structured Dirichlet smoothing (SDS) model while the second scenario is to modify
the parameters used in the model as in the proposed Adaptive Dirichlet smoothing
(ADS) model. The third stage of the proposed framework involves the proposed SQE
method to enhance the retrieval performance of DROs by improving the quality of
candidate terms that are added semantically to the entire query term. Extensive experiments were conducted to evaluate the effectiveness of the proposed methods,
model and IR framework using the publicly available CHiC2013 collection. The
experimental results show that the performances of the proposed EDE method, ASDS
model, SQE method and IR framework improve by 10.5%, 11.3%, 8.1%, and 25.7%
(mean average precision measure) respectively over conventional methods, models
and frameworks.
Description
Keywords
Computer science