Information Technology Reference
In-Depth Information
there was no significant interaction between the factors. In recent studies, the
search performance is often assessed on the basis of the highly ranked items such
as the top 10 results, and high recall is not always the standard assumption in
the assessment of search results [11]. Therefore, regardless of no significant im-
provement in the recall, it is noteworthy that the precision has been significantly
improved in this study.
In our auto-suggest interface, when a term is selected from the suggestions, a
list of school documents are displayed as candidates (Figure 3). Currently, such
candidates are listed in the order of document titles
8
, and sometimes it is not
straightforward for the users to find out a desired item from a lot of candidates.
A previous study investigated the ways of organizing items suggested by auto-
completion in a known-item search task [19]. It was concluded that alphabetical
order was suitable for a domain-independent lexical thesaurus, while domain-
dependent grouping meaningful for users should be chosen carefully because
every grouping strategy did not improve the search performance [19]. Further
research will be needed for the organization of the suggested items, so that the
quality of the search results can be elaborated.
6 Concluding Remark
In this paper we demonstrated the potential effectiveness of the auto-suggest in-
terface, which exhibited better performance when used with domain vocabulary.
The domain vocabulary utilized in this study was developed manually by school
teachers and translation staff members. The effectiveness of the auto-suggest in-
terface, therefore, may be dependent on the quality of the domain vocabulary. It
is important to further investigate ways of developing vocabulary with semantic
relations, so that the dynamic term suggestion technique can be applicable to
the broader context other than the school education domain.
It would also be necessary to continue the evaluation with real users of mul-
tilingual school documents because the experiment presented in this paper was
conducted with undergraduate students. In addition, currently the auto-suggest
interface is integrated with the search box of the portal, but not fully integrated
with the faceted search. We are planning to further investigate the effective-
ness of the auto-suggest interface in cooperation with real users by using the
full-featured faceted search system.
References
1.
http://www.mext.go.jp/b_menu/shingi/chousa/shotou/042/houkoku/
08070301.htm
(in Japanese, accessed 2010-03-18 )
2.
http://www.mext.go.jp/b_menu/houdou/21/07/1279262.htm
(in Japanese, accessed 2010-03-18 )
8
In the order of Japanese phonetic characters, which corresponds to the alphabetical
sequence in English.
