Why do we need RDF trust ?
- Anyone can write RDF description.
- There are trusted RDF Information and not trusted RDF Information.
- The trust of RDF information depends on the user.
- Depends on if the user believes the RDF information
is true or not.
- Depends on if the user believes the RDF information
is important or not.
(Importance of RDF
-- similar to the concept of Google PageRank.)
Describing RDF trust in RDF
- By describing RDF trust in RDF,
it can be processed automatically.
(The greatest merit of describing information in RDF)
- Determine the value of RDF trust with the URI in which the RDF
description was stored (provenance URI).
Prototype: Search Result Filter by Shared RDF Bookmark
- Search results from general search engine contains much noises.
- Thus, we have to pick up the URI among many URIs.
- By filtering those search results with shared bookmark information,
more sophisticated search results can be provided.
- In this prototype,
I tried to make those information collaborate automatically
with providing all of those information in RDF.
That is,
"Attach trust information lead from bookmark information provided
in RDF to the search result which is also provided in RDF and
present the result to the user."
Functions Implemented in This Prototype
Following functions are already implemented in this prototype:
Google search result RDF converter
Parse Google search result and convert them into RDF so that
they can be processed by the Trust Leveling Engine.
RDF Bookmark Parser
Parse the user's RDF bookmark and the other user's bookmark
which is bookmarked by the user's bookmark.
Trust Leveling Engine
Give trust level based on Google search result in RDF and
RDF bookmark and show the result to the user.
The Trust Model in This Prototype
To store trust information, I added following 2 attributes:
b:retrievedFrom
The URI in which the bookmark information was stored.
b:prefer
The preference of the bookmark (prefer or deny).
Collaboration of Search Engine and Shared RDF Bookmark
- The Search Result from the search engine is ...
"URIs concerned to the given keywords, generated
by the search engine."
- The Bookmark of the user is ...
"highly trusted URIs generated by the user."
- The actual function of this system is filtering the search result.
But its definition is:
"Automatic collaboration of search result provided in RDF and
bookmark information provided in RDF, based on the information of
their origin (provenance URI)".
Legacy Search Engine System
- Unable to give feedback to the search result.
- User cannot access to the trust data in the search engine
(e.g. "PageRank" data in Google).
- Trust by search engine and trust by user (bookmark) are stored
separetely.
- Thus, user should maintain those information by hand.
Search Result Filter by Shared RDF Bookmark
- Can give feedback to the search result by modifying RDF bookmark.
- Can re-use other users' trust (RDF bookmark).
- The trust data in the search engine and the trust by
users (RDF bookmark) can collaborate togather.
Prototype Demo (if time allows)
is available in
http://iggy.w3.org/shiraishi/bookmark/google+bookmark.html
Demo Screenshot 1 (Google Query Result)
Google Query Result with keyword "Boston" and "Weather".
Demo Screenshot 2 (Query using Shared RDF Bookmark -- prefer)
Query result with "prefer" bookmark.
Demo Screenshot 2 (Query using Shared RDF Bookmark -- deny)
Query result with "deny" bookmark.
Collaboration of This Prototype and Annotea Shared RDF
Bookmark Function
- This prototype can parse the Annotea shared RDF bookmark.
- Amaya acts as the authoring tool of the shared RDF bookmark.
- This prototype is one of the applications that use
the Annotea shared RDF bookmark.
Future Works
- Apply this mechanism attaching trust information to general RDF
description but search result.
- Construct distributed RDF database for translations of
the Web documents (e.g. recomendations, notes, etc.) and
attach trust to them.
- Improve the implementation of this prototype.
- Enhance parsing RDF function by introducing general RDF
processing library.
- Take keyword/topic of the bookmark item into count for
leading RDF trust.