Optimization of Semantic News Feed Query
No Thumbnail Available
Date
2013-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
RSS feeds are text-content rich, semantically heterogeneous, dynamic XML element and
streamed in an asynchronous and pull strategy. Hence, for efficient retrieval of RSS feed,
window-based and semantic-aware feeds querying operators has been proposed recently. It is
shown that the use of semantic information improves the relevance of query result at the cost of
degrading the efficiency of the system. To benefit from query execution on semantic information
while keeping the efficiency of the system, we have proposed multi-query optimization strategy
and semantic load shedding technique. The proposed MQO strategy accepts multiple semantic
news feed queries and examines the relationship that exist between each queries with respect to
their defined window and their semantic similarity. Then, a MQE chain is generated based on
the relationship examined for the efficient manipulation of queries at run time. And each time a
new query arrives, it is added to its appropriate chain. On the other hand, the proposed semantic
load shedding technique drops non-relevant data elements from a shared window and generates
a new window with optimal size. The approach first extracts a set of representative keys from
multiple query‘s search term and then the keys are enhanced semantically with concepts
retrieved from knowledge base. Then before Multi-Query execution, using the enhanced key set,
the news items in the shared window are classified into relevant and non-relevant classes based
on their key content. Those elements in the relevant class is directly chosen to be a member of
the new window and among the elements in the non-relevant class a representative sample is
selected at random with a controlled sampling precision level to be included into the new
window. The resulting new window contains reduced data elements and yet satisfies the need of
all queries in the given MQE chain. The approach is implemented as an operator called
quickDrop operator. The operator quickDrop (W, T, K, e) takes four parameters; W
denotes a shared window among a set of queries for Multi-Query execution, T denotes the
attribute of the news feed for shedding i.e. title or description, K denotes a set of enhanced keys
and e denotes the level of precision to be used for sampling and it returns a window with optimal
number of elements. Coordinated with the MQO approach, the proposed semantic load shedding
technique performs efficiently for the performance increase of the system. To validate the
proposed approach, we have developed a prototype and experimental results show a significant
performance increase.
Key Words: SNF Query, NAT, Windows Boundary Similarity Measure, MQE Chain Generator,
Window Relationship, MQE Rules, Semantic Load Shedding, QuickDrop Operator.
Description
Keywords
Snf Query; Nat; Windows Boundary Similarity Measure; Mqe Chain Generator; Window Relationship; Mqe Rules; Semantic Load Shedding; Quickdrop Operator