Information Technology Reference
In-Depth Information
For brevity, in this chapter we did not presented all the experiments we have
conducted in respect to explore the behavior of the designed ρ -index. For ex-
ample we tackled the issue of the parameter settings modification also at the
upper levels of the indexing structure and their impact on the particular ρ -index
creation time and consequent search complexity.
Hence, to this relates also our next future work. Firstly we want to carry out
more tests to be able to precisely predict the ρ -index properties under certain
parameter settings and hence to be able to find optimal settings for the testing
data. Afterwards, we would like to carry out tests on more testing data in order
to investigate the scalability of the ρ -index in respect to the number of vertices
and to the number of edges in the indexed data. In our near future work we would
also like to implement the algorithm for discovering all connections between two
vertices as the ρ -index allows such utilization according to [7].
As the experiment on the real life data shown it is interesting to study complex
relationships between publications in the citation graph in order to help the user
identify important publications in the particular scientific field. We presented a
search for all paths between two publications where the result was presented
as network ordered according to the date of publication. But also another in-
formation retrieval techniques could be used to rank the search result. We also
identified possible new approaches using our indexing structure like a search for
a relation between one reference publication and a set of core publications to
improve the accuracy of the search.
The future work concerning the information retrieval part of our work is to con-
firm the contribution of a complex relation search in the citation graph. In the first
place, we would like to verify using standard information retrieval techniques the
relevance of the publications recommended by the system. Secondly, we want to
study the relevance to this problem of using a set of core publications instead of
just one single publication in the search. Last but not least we also want to incor-
porate and evaluate the search for a connection between two publications into the
system because it provides a notable alternative to the indirect citation relation.
References
1. Agrawal, R., Borgida, A., Jagadish, H.V.: Ecient management of transitive re-
lationships in large data and knowledge bases. In: Proceedings of the 1989 ACM
SIGMOD international conference on Management of data, pp. 253-262. ACM
Press, New York (1989)
2. Agrawal, R., Dar, S., Jagadish, H.V.: Direct transitive closure algorithms: design
and performance evaluation. ACM Transactions on Database Systems 15(3), 427-
458 (1990)
3. Agrawal, R., Jagadish, H.V.: Hybrid transitive closure algorithms. In: McLeod, D.,
Sacks-Davis, R., Schek, H.-J. (eds.) Proceedings of 16th International Conference
on Very Large Data Bases, Brisbane, Queensland, Australia, August 13-16, 1990,
pp. 326-334. Morgan Kaufmann, San Francisco (1990)
4. An, Y., Janssen, J., Milios, E.E.: Characterizing and mining the citation graph of
the computer science literature. In: Knowledge Information Systems, vol. 6, pp.
664-678. Springer, New York (2004)
 
Search WWH ::




Custom Search