2006 Fiscal Year Final Research Report Summary
Development of an answer question system that reuses the existing help manuals as knowledge resource written with language
Grant-in-Aid for Scientific Research (C)
|Allocation Type||Single-year Grants |
|Research Institution||Ochanomizu University |
KOBAYSHI Ichiro Ochanomizu University, Faculty of Science, Associate Professor -> お茶の水女子大学, 理学部, 助教授 (60281440)
|Project Period (FY)
2004 – 2006
|Keywords||question-answer system / tailor-made information providing / language instruction / application software operation / help text / user manuals / subject expriment / text summarization / generation|
1 Knowledge acquisition and new matching method
(1)As knowledge resource for our question and answer system, we use not only help manuals of application software, but also use knowledge obtained from Web.
(2)In order to flexibly provide appropriate information to users, we proposed a paraphrasing dictionary that connects user's ambiguous expressions to a keyword for information retrieval.
2.User model construction
(1)A user model was constructed based on a subject experiment in that the patterns of users' behaviors of operating application software were observed. A user model, which can provide appropriate help information to a user, was constructed.
3.Verbal operation of application software
(1)We constructed a prototype system that can operate application software, MS Outlook, MS Word, MS Excel in this study, through user's language instructions.
(2)We constructed task and domain ontology of application software, referring to use's manual of application software operation. And based on the ontology, it has been successful that the proposed system can be operated by adequately interpreting user's language instruction.
4.Proposal of intelligent information providing to users
(1)We proposed a mechanism that provides tailor-made information to a user corresponding to the user's interest.
5.Enhancement of the functions of the prototype system mentioned at 3 (1)
The prototype system mentioned at 3 (1) was improved in terms of operation ability. Especially, we introduced domain and task ontology obtained at 3 (2) into the system.
6.Evaluation of the proposed methods
In order to evaluate the proposed mechanism, we compared the output produced by the system and humans through a subject experiment. As the result of the experiment, we figured out that the system can provide proper information to users with the same ratio as humans can do.
Research Products (22 results)