Tabulator: Exploring and Analyzing linked data on the Semantic Web

1. Tabulator:Exploring and Analyzing linked dataon the Semantic Web Tim Berners-Lee, Yuhsin Chen, Lydia Chilton, Dan Connolly, Ruth Dhanaraj, James Hollenbach, Adam Lerer, David Sheets Decentralized Information Group, Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, USA. 2008-10-01 Presentation by JongHeumYeon, IDS Lab.

2. Contents • Introduction • Methodology • Scenarios • User Interface • Web vs. Semantic Web/Explore vs. Analyze • Exploring in outliner mode/Views of RDF data • Network access algorithms • From representation to RDF/What to dereference/Hash signs and redirects • Friend-of-a-friend conventions/Inference on the client • Related Work • Future Work • Evaluation • Conclusions

3. Introduction • The Tabulator • RDF browser • For users • To access and interact with the entire web of RDF data • For developers • To post their data in RDF, refine and promote RDF linking standards, • For providers • To see how their data interacts with the rest of the Semantic Web. • demonstrate and utilize • the power of linked RDF data with a user-friendly Semantic Web browser • A challenge for Semantic Web browsers • domain-specific applications in a generic program

4. Motivation • HTML • could immediately see the results of the work • add immediate value to his page by linking to other chosen resources • URI • Many projects such as Biopax are currently available online • but consist of a small number of huge archive files • the emphasis on inference over fixed datasets • lack of linked data is self-sustaining • without things to link <-> no incentive for putting one’s own data on the web • lack of a straightforward generic data browser • give an immediate feedback and gratification of online linked data.

5. Methodology • The browser should be as easy as possible for a new user to pick up • and easy for developers to extend with their own ideas • Asynchronous Javascript and RDF (AJAR) platform • HTML DOM • HTTP • XML • RDF • SPARQL

6. Scenarios The documents published by W3C and its organizational structure are on separate web pages. As a working group chair, Alice wants a list of documents which are in Last Call status, their editors, and the email address of the domain leaders responsible for the working groups for which the document is a deliverable. She sorts them by date comments are due, and finds one she almost missed. She emails the chair of the working group asking for a extension of the deadline. Bob has no idea what a line for $364 on his credit card statement is about. He looks at it on a calendar view, sees it was a Saturday, but still doesn’t know. He looks for photos he took at the same time, and displays it on the same calendar. Now he understands: he took the kids to an amusement park.

7. Scenarios Charlie notes that Danielle’s blog is critical of his work, and wants to understand why that might be. Danielle has in her data file information pointing to papers she has written, projects she has worked on, and professional acquaintances. Charlie realizes they have an acquaintance in common who can probably help resolve the issue. An existing relational database such as the W3C internal enterprise database, designed without semantic web export in mind, is exported into linked RDF. The Tabulator is found to be effective for browsing data in the database, both for the original users of the data, and also for new users.

8. User Interface • Web Brower • navigates along links between documents • Semantic Web Browser • navigates along relationships (predicates) in a web of concepts • have an awareness of the underlying web of documents (and query services) • Tabulator • Primary : the graph of logical information • Secondary : the web of documents • The user explores an abstract web of data • conjunction of all the graphs of documents that have been read • Users check the provenance or source of any piece of information • data is not yet accessed (blue), already fetched (green), in progress (yellow), or failed (red)

9. User Interface http://dig.csail.mit.edu/2005/ajar/release/tabulator/0.8/tab.html http://dig.csail.mit.edu/2007/wiki/Projects.rdf#OpenLinkedDataProject

10. User Interface • Explore vs. Analyze • WWW is something we explore, following links on a hunch • Exploring the semantic web in this way involves • moving from node to node • finding more arcs • reassessing our next move • Tabulator • operates in two interlocking modes • Exploration • Analysis

11. User Interface • Exploration mode • explore an RDF graph in a tree view • Analysis mode • select certain fields (arcs or predicates) to define a pattern • ask the Tabulator to find all examples of that pattern • match the query pattern to the RDF graph

12. Analysis mode

13. User Interface • Outliner mode • People are typically very comfortable in a tree-like environment • Much data in the world has been organized into trees • The web is largely composed of overlapping trees with local roots all over the place

14. Views of RDF data Map View

15. Views of RDF data Calendar and Timeline View

16. Views of RDF data SPARQL: Formulating Queries in Tabulator

17. Network access algorithms • The dereferencing of links in RDF has not been explored to a great extent to date • If an inference engine simply downloads data • whenever it comes across a new URI • in an unbounded open web • continue without limit • In the case of a user browser • there is direction • in the direction the user chooses to navigate • in a specific query

18. Network access algorithms:Limitations • Server has to anticipate what data will be useful • In the general case of an exploring user • Some data would be just too numerous • e.g. for a person, an entire track of every known date/location • Not all data is available in machine-readable form

19. Network access algorithms:Hash signs and redirects • URI containing a hash (#) • http://dig.csail.mit.edu/2005/ajar/ajaw/data#Tabulator • http://dig.csail.mit.edu/2007/wiki/tabulator#project • Information on many things in the same file • URIs without a # if the URI denotes an arbitrary thing • Server responsd : HTTP 303 • Forwarding the client to a document that contain information

20. Network access algorithms:From representation to RDF • Conventions - should the whole set be a standard? • Content-type is definitive • Text/plain: Guess with a warning • application/rdf+xml Parse as RDF/XML • text/rdf+n3 Parse as N3 (*not imp.) • text/html • link rel=meta in HTML • GRDDL profile • RDF/a ?? • *+xml*: • GRDDL profile

21. Network access algorithms:From representation to RDF

22. Network access algorithms:What to deference • If user is browsing information about a subject x • Looking up involves • looking up the URI of x itself • looking up any y where the store includes the fact that { x rdfs:seeAlso y}.

23. Network access algorithms:Friend-of-a-friend conventions • FOAF, given by rdfs:seeAlso, is a link to their FOAF file • The protocol is • to load the resource linked by rdfs:seeAlso • to merge (’smush’) nodes with the same mailbox or mailbox hash • As these are inverse-functional properties • dealt with by the inference layer

24. Network access algorithms:Inference on the client • The limited inference performed • <owl:sameAs> smushing • Merging nodes <owl:sameAs> each other • Merging of equal nodes • Merging nodes • With identical functional or inverse functional properties • Essential for the FOAF convention • rdfs:subPropertyOf • used for finding subproperties of rdfs:label for the user interface • URI canonicalization Certain URIs • have only syntactic differences and are always equivalent • Hashed property entailment • { ?p :Sha1Property ?q. ?x ?p y?. ?y crypto:sha1 ?z} => {?x ?q ?z} • This is needed for matching foaf:mailbox to foaf:sha1 mbox. • OWL-DL entailment is not supported.

25. Future Work • Extension of query-by-example to intuitive rule-building systems • Experiments with a variety of remote stores & query engines • Provide a more graphical user interface using Scalable Vector Graphics (SVG) or 3D • Extensions of the Fresnel language to cover input form views are one possibility • user may want to add from the existing data • client save back a modified file, share every change over the network with a server, other clients viewing or editing the same information

26. Evaluation W3C staff members test and evaluate No way to undo or go back to a certain previous state Lack of cognitive map Requirement of spreadsheet like navigation

27. Conclusions • Domain-specific applications • will always be important • will always do better at specific tasks than the general one • Interoperability between a generic client and an application-specific one is crucial • In order to enable generic browsing • users should be encouraged to leave sufficiently powerful user interface tips in ontologies • generic application can acquire the ability • to provide an effective and useful interface to data from previously unknown domains

28. Current Works in2007 Tabulator Redux: Writing Into the Semantic Web, unpublished, 2007 Object Selection Predicate Selection Editing in Table Mode Network Protocol for Writing

29. Future Work in2007 Browser integration Updating Information Collaboration Predicates Social Policy UI/Usability Longer term devlopments

30. Interesting Points • Written by Tim Berners-Lee • Scenarios in semantic web • Give examples of practice use case off semantic web • Provide a great vision of semantic web browsers