Unifying Reasoning and Search Based on User Interests (I-ReaSearch)
(Joint efforts by WP2, WP4 and WP7a)
Participants and Responsibility
Participants |
Affiliation |
Responsibility |
Yi Zeng, Yan Wang, Cong Wang |
WICI |
Building up the theory, initial implementation using real dataset(Currently SwetoDBLP and Medline are used), LarKC plugin development. |
Vassil Momtchev |
?OntoText |
Integrating initial implementation into LLD and provide user interface based on LLD. |
Bo Andersson |
?AstraZeneca |
End User Evaluation and feedbacks. |
A Brief Overview
The ?ReaSearch approach proposed in [Fensel & van Harmelen 2007] is aimed at meeting the barrier of scalability for Web-scale reasoning. It's core philosophy is to select an appropriate subset of semantic data for reasoning and is trying to solve the scalability issue by incomplete reasoning since the dataset on the Web itself are incomplete anyway. The criterion and concrete methods for selecting a good subset is one of the main task in the LarKC project.
Our main efforts are on the direction of context-aware approaches, which emphasize the power of user interests during the selection process and how it can be used as a factor to provide good selected subset for the querying and reasoning process. Following the notion in [Fensel & van Harmelen 2007], we title the efforts as "I-?ReaSearch").
In the I-?ReaSearch framework, user interests are evaluated from various perspectives [Zeng 2009]:
- Retained Interests
- Cumulative Interests
- Interests Lasting Time
- Interests Appear Time
When combined with concrete strategies of I-?ReaSearch, each of them can produce a unique set of results to meet a specific user needs. It is currently composed of two concrete strategies, namely:
- User Interests Based Query Refinement
- Interleaving Selection and Reasoning Based on User Interests
User Interests Based Query Refinement
For the strategy of user interests based Query Refinement, it adds more constraints to the user input query according to user interests extracted from some historical sources (such as previous publication, visiting logs, etc.). We emphasize that this approach does not select the subset for querying in advance. Instead, it utilizes the user context to provide a rewritten query.
Status
- Plugin available, with demo using the SwetoDBLP dataset. (WICI)
In the Process of Integrating it to LLD and create the first prototye based on LLD. (?OntoText, WICI) Current version of SPARQL query on LLD supports the proposed weighted interests well, and it is not far to integrate interests into the original query.
The prototype of "Context-aware Linked Life Data Search" is available at http://www.wici-lab.org/wici/context-aware-LLD/
- End User Evaluation Pending (Astrazeneca)
Interleaving Selection and Reasoning Based on User Interests
For the Strategy of interleaving selection and reasoning based on user interests, it emphasize a selection step before querying and reasoning on the data, since user interests might help to find a more relevant sub dataset for each specific user compared to querying on the whole.
Status
- Plugin available, with demo using the SwetoDBLP dataset.
- More detailed Explanation will be available in D2.3.2 and D4.3.2.
Related Deliverable
- D-2.3.1 [submitted]
- D-2.3.2 [submitted]
- D-4.3.1 [submitted]
- D-4.3.2 [submitted]
References
[Fensel & van Harmelen 2007] Dieter Fensel and Frank van Harmelen. Unifying reasoning and search to web scale. IEEE Internet Computing, 11(2):96, 94–95, 2007.
[Zeng 2009] Yi Zeng, Yiyu Yao, Ning Zhong. DBLP-SSE: A DBLP Search Support Engine In: Proceedings of the 2009 IEEE/WIC/ACM International Conference on Web Intelligence, IEEE Computer Society, Milan, Italy, September 15-18, 2009.