USF Department of Computer Science Peer-to-Peer Knowledge Sharing David Wolber.

USF Department of Computer Science

Peer-to-Peer Knowledge Sharing

David Wolber



Associative Thinking Assistant


Motivation

• Desktop is outdated

• Digital libraries abound, but no tools (INFORMATION OVERLOAD)

• Avoiding a Googleopoly

• Peer-to-peer is a great new model, but sharing knowledge is hard.


Sneak Preview

• Develop webtop, not desktop tools

• A common search API and registry

• Zero-input publishing

• Open PC Movement


Today’s Desktop

BookMark File Manager Hyper-LinksManager

WebSphinx (Miller, Bharat)


Desktop tools are outdated

• Filing cabinet metaphor no longer fits

• Unnecessary barrier between local system and web

• Computer is a communication tool, a node in the knowledge base of the world


I see the documentsWhere are the people?


Motivation






Information Sources

• Google, Amazon, etc.

• History: Way-back machine

• Technorati– The World Live Web

• Domain Specific:– ACM Digital Library– Citeseer– MLA for literature


www.technorati.com


CiteSeer: http://citeseer.ist.psu.edu/


ACM


Internet Archive, Wayback


They all provide similar functionality, but…


Motivation






Media Conglomeration

• FCC lessening anti-trust laws

• Most media now owned by a handful of companies

• How many of us go through Google to get our information.

• Can we trust Google to stay “good”?


Motivation






Our Virtual Brains are modeled after our real ones


Research Tasks

Communication

Email

Blog

IM

Web page

Knowledge Creation

Bookmark

Create note/doc

Put in folder

Add link


The Big Question

How much of the information hidden

within your personal web is hidden due

to privacy concerns?


Associative Thinking Agent

Personal Personal Personal General-Purpose Search Engine

Specialized Search Engines

Associative Thinking Agent

Figure 1. The proposed system allows any entity that can provide associative data to interested parties to register as associative thinking agent source. This includes general-purpose search engines like Google, specialized search engines like the Internet Archive, and ordinary users of personal computers. Associative thinking agents, residing on personal computers, help the user identify sources and information relevant to the user’s current task.

Source list


Personal Metadata

Cassini server

PW Service

WebTop

Client

Personal Computer 1

File System

Personal Metadata

Cassini server

PW Service

WebTop

Client

Personal Computer 2

File System

Assoc SourceUDDI Registry

WTGoogle

WTTechnorati


WebTop Client

• Tree View of results

• Information processing on node expansion

• Choose from dynamic list of sources

• N-Degree Heterogeneous Associations


Associations

• Outward Links (hyperlinks)

• Inward Links

• Similar Content Links

• Collaborative filtering Links


More Associations

• General Research: AuthorOf, collaborator, co-cited, home page of,…

• Domain-Specific– Films: Acted in, Directed– Campaigns: Gave money to


Separate Association Types

Inward Links Browser Outward Links

Content-Similar


IOC Tree

IOC Tree

I --- C – O --I

Browser


Associative Source API and Registry

• A common search/link info API

• A central UDDI registry

• Allows for meta-search clients with dynamic lists of sources.

Clients can access even those sources created after the client itself.


Organizational Sources

• Web Service Wrappers

• Wrappers implement one or more of the Associative Source API methods

WebTop Client

WTGoogleAssoc Service

Google Web Service

USF Google


Scenarios

• Search within the boomarks/links/papers of an expert

• When open one of your own papers, see newly created documents that cite it, that have similar content

• Easily navigate paper-author trail


The Personal Web

2nd Degree

1st Degree

Nth Degree

PersonalWeb

Personal Space consists of highly interrelated files, i.e. a personal web

The Web should be personalized to each user, i.e., a personal web.

Personal Web-- local documents and “bookmarks”

N-degree Neighborhood– can be reached with n clicks.


You can be a Search Engine

• Personal Web Analyzer– In/out links between documents– Full-text index– Characteristic words– Person/proper names in document

• Access between personal webs– Automatically installed mini-server– Auto-registration of web service


Connecting Personal Webs

People create knowledge-- bookmarking, annotating, linking, synthesizing-- every day.

Much of it is never communicated


Future Work

• Automated source discovery

• Alternative page ranking– Source reputation measures– In-links from groups of people, e.g.,

CS Professors


Get Involved

• http://webtop.cs.usfca.edu

• [email protected]

http://webtop.cs.usfca.edu/

mailto:[email protected]


WIKIs


Lawrence Lessig’s Blog

• http://www.lessig.org/blog/


Web Services

• Programmatic interface to web data

• Software agent access to data

• Helps separate View/Model

• More robust than “web scrapers”

• XML/Soap

• Rest-ful Interfaces


Web Services

• Web services allow clients to aggregate information from various sources, process, then display in a custom fashion.

• More robust than scraping web pages

client

source

source

GUI


XML/Soap Web Services

• XML/Soap

• WSDL

• UDDI


Know more about yourself

Analysis of file system– Linking between documents– inward links– characteristic words of documents– what people are referred to– what web pages

USF Department of Computer Science Peer-to-Peer Knowledge Sharing David Wolber.

Documents

web computer

world slide

literature slide

bharat slide

wayback slide

link slide

real ones slide

googleopoly peer