1. Exploring Personal CoreSpace For DataSpace Management Li Yukun and Xiaofeng Meng WAMDM Lab Renmin University of China

2. Outline  Introduction  CoreSpace Overview  CoreSpace Design  CoreSpace Implementation  Conclusion

3. Motivation Query Find a pdf file I downloaded from a web page and stored in a directory of D drive. Revisit a picture I developed for MDM2008 one years ago. Background With increasing of personal data set, PIM becomes a serious problem and a hot research issue; The current tools can not work well in some cases.

4. Related work  Current solutions Traditional tools  Folder explorer, Desktop Search DataSpace Support Platforms (DSSPs)  Personal data integration (Xin Dong,etc.)  Association-based query ( Salles MAV, etc. )  Data Resource Model RSM, SLN (Hai Zhuge, etc.)  Our solution Based on user features  Users play a key role  Revisit is an popular access style Research focuses  Highlight the role of users  Produce an effective approach for exploring PDS

5. Problem Definition Personal DataSpacePersonal CoreSpace Classify Exploring -Modeling user features -Exploring based on user features

6. Contributions  Propose CoreSpace Model Divide the semantic links among PDS into two classes:  Objective Semantic Link(OSL)  Memory-based Semantic Link(MSL) Describe Personal CoreSpace(PCS) based on Resource Space Model (RSM).  An ontology of Personal CoreSpace Discover several types of meaningful MSLs Design an ontology of PCS based on the MSLs  A facet-based search interface of PCS Propose a method to translate the PCS ontology into a facet-based search interface. Validate the effectiveness of our methods by implementing a prototype system.

7. Outline  Introduction  CoreSpace Overview  CoreSpace Design  CoreSpace Implement  Conclussion

8. Features of personal data  Features of personal data Versatile, heterogeneous, personalized, complex, evolutionary  Features of personal data operations Pay-Go Integration Known-item relocation-- “revisit” Multiple query methods Simple interface

9. Resource Space Model  A resource space is a n-dimensional space Axis : Xi is the name of an axis. Xi = (Ci1;Ci2;...;Cin) represents an axis with its coordinates and the order between them. Coordinate: C denotes the coordinate name in form of a noun or a noun phrase. Point: determines one or a set of entities, we denote it as PCS(X1;X2;...;Xn).  Data operation [1] H.Zhuge, Communities and Emerging Semantics in Semantic Link Network: Discovery and Learning, IEEE Transactions on Knowledge and Data Engineering, vol.21, no.6, 2009, pp. 785-799. [2] H. Zhuge. Resource space model, its design method and applications. The Journal of Systems and Software 72 (2004) 71-81. [3] H.Zhuge, The Web Resource Space Model, Springer, 2008.

10. Personal CoreSpace Model  Personal DataSpace Data item  Attribute Owner Relationship  Personal CoreSpace A n-dimensional space Axis : Attributes of personal data items. Coordinate: Values of a certain attribute, which can be a tree structure. Point: A personal item or a set of personal items.

11. Outline  Introduction  CoreSpace Overview  CoreSpace Design  CoreSpace Implementation  Conclussion

12. Personal CoreSpace Ontology  Two type of attributes Natural attributes  Name, Type,Access time, Directory, Size, Source User-based attributes  Access frequency, access type, related task

13.  Type: {Email, Web pages, Picture, Documents,…}  Access time {”Today”,”Yesterday”,”Last week”,”Last month”,”Last year”,”One year ago”}  Directory A Tree structure  Size {(0,10K]; (10K,100K]; (100K,1M]; (1M,10M]; (10M,-)}  Sources {Self-developed, Cloned}  Access frequency {(1,5]; (6,15]; (16,50]; (50,-]}  Access type: {Read-only, Modified}  Related tasks A personal task set Personal CoreSpace Ontology

14. Outline  Introduction  CoreSpace Overview  CoreSpace Design  CoreSpace Implementation  Conclussion

15. CoreSpace Implementation  System Framework User behavior monitor Storage agent Item identify agent Query processor  Features PayGo evolution From CoreSpace to facet search Extendability

16. From CoreSpace to facet search  Method Take each coordinate Xi as a facet Fi, and take its coordinates as the options of facet Fi. Based on the hierarchical structure of PCS, we can easily construct a facet-based search interface.  Facet-based query logical Let X and Y’ be two selected nodes of facet tree, and they can be regarded as two conditional expressions. Our method is detailed as below.  If X is parent of Y, it means X and Y = Y;  If X is brother of Y, it means X or Y;  If X and Y are neither parent relation nor brother relationship, it means X or Y.

17. An example of query algebra The red nodes represents those options selected by user According to the rules we can get the logical expression R = {Xi | (Xi. type = JPG ∨ Xi.type =VSD) ∧ Xi. place = ”D : \Picture”}

18. Implementation

19. Outline  Introduction  CoreSpace Overview  CoreSpace Design  CoreSpace Implementation  Conclussion

20. Conclusion  This is just a preliminary work Propose a CoreSpace model Propose a method to explore PDS based on CoreSpace  Future work Try to discover more rules of user memory Enrich the ontology of PCS

21. Thanks