Page 143 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42

142 Journal of Library Science in China, Vol. 8, 2016

The evaluation results show that the four usage modes of terminology services obtain consistent
results in the two information retrieval systems – OPAC and Baidu. The usefulness rank of the four
usage modes is: 1) multiple choice for query expansion (Mode 1), 2) hybrid of query expansion
and replacement (Mode 3), 3) automatic query expansion (Mode 4) and 4) single choice for query
replacement (Mode 2). In effectiveness, Model 1 and Mode 3 are superior to Mode 2 and Mode
4 because users have a greater deal of freedom in term selection in Model 1 and Model 3 and
thus are able to form search queries which can better reflect users’ information requirement to
improve search results. The effectiveness of Model 3 (3.92, 3.85) is less than or equal to that of
Model 1 (3.92, 4.04), which means the function of replacing original query term does not play any
effect. In efficiency, Model 4 is the best because the system can complete the search automatically
without the need for users to do any work, Model 1 is the second best, and the rest two have no
big difference. To our surprise, in the design of Model 2, we intentionally combined term selection
with query execution together to reduce users’ clicking times and improve operation efficiency.
However, the results indicate that this practice did not help too much in fact. In user satisfaction
with the highest weight, Model 1 and Model 3 are much better than Mode 2 and Model 4. In learn-
ability, Model 4 is a little greater than or equal to Model 1, both outperforming Model 2 and Mode
3 whereas Model 2 is the worst. On the whole, Model 1 (i.e. multiple choice for query expansion)
is the best usage mode which obtained the highest or the second highest scores in effectiveness,
efficiency, user satisfaction and learnability and achieved the best in the overall usefulness,
whereas Model 2 (i.e. single choice for query replacement) is the worst usage model, which is not
effective, not easy to learn, and have no much improvement in efficiency.

5 Conclusion and future work

In this paper, the construction of terminology services was first studied. Two important modules
in a terminology service system, i.e. term storage & retrieval and terminology services, were
investigated with the use of Semantic Web technologies and REST Web Service architecture.
By taking Chinese Thesaurus in SKOS/RDF as a source vocabulary, six basic terminology
services, including SearchConceptByKeyword, getBroaderConcept, getNarrowerConcept,
getRelatedConcept, getSynonym, and getEnglishTranslation, were built. Next, this paper focuses
on investigating the application of terminology services in information retrieval systems. On the
one hand, taking the GetSynonym service as an example, we evaluated the usefulness of this
service in the library OPAC system and Baidu Search Engine. The evaluation results showed
that the F value of OPAC increases 13% whereas the P@5 precision of Baidu increases 16%. On
the other hand, we designed four usage modes of terminology services in information retrieval
systems and evaluated their usability. The evaluation results showed that multiple choice for query
expansion was the best usage mode.
However, the terminology services built in this study were only based on a single vocabulary and