Transcript
Information & Knowledge Management
Discussion of “Collective Knowledge Systems: When the Social Web Meets the Semantic Web”
from Tom Gruber (TomGruber.org)
Marielba ZacariasProf. Auxiliar DEEI
FCT I, Gab 2.69, Ext. 7749
mzacaria@ualg.pt
Summary
The Vision of Collective Intelligence
Collective Knowledge Systems
The Role of the Semantic Web
Augmenting User-Contributed Data with Structured Data
Enabling Data Sharing and Computation Across Applications
Example
Collective Knowledge System for Travel
The vision of Collective Intelligence
Web 2.0 (Social Web)
Class of web sites and applications in which user participation is the main driver of value
Wikipedia, MySpace, YouTube, FIicker, Del.icio.us, Facebook, Technorati, etc., Blogger, WordPress
The vision of Collective Intelligence
web 2.0web 1.0
The vision of Collective Intelligence
Harnessing Collective IntelligenceHyperlinking works as brain synapsisYahoo!’s role as a portal of net users’ collective workGoogle’s PageRank search exploits web structure rather than just doc characteristicseBays’ product is the collective activity of its usersAmazon has made a science of user engagementFlicker & Del.icio.us pioneered folksonomies Wikipedia based on the idea that any user may edit any entryCollaborative spam filtering like CloudmarkGreatest internet successes driven by viral marketingInternet infrastructure (php, apache, mysql, python) mostly based on peer-production of open source software
The vision of Collective Intelligence
Collective Intelligence or Wisdom of the Crowds
Value created by collective writing articles in wikipedia, sharing tagged photos in flicker, sharing bookmarks in del.icio.us or streaming their personal blogs in the open space called the blogosphere
Unmatched potential for knowledge sharing
Collected intelligence
But not collective intelligence
No emergence of new levels of undersanding of knowledge
The vision of Collective Intelligence
Collective intelligence has been goal of several visionaries
Grand challenge is to boost the collective IQ of organizations and society
human-machine system for
collecting knowledge for learning
evolving technology for collective learning
humans and machines actively contribute doing what they do best
The vision of Collective Intelligence
Tim Berners-Lee inventor of the semantic web
Semantic web is an extension of social web in which information is given precise meaning
better enabling people and computers to cooperate
The vision of Collective Intelligence
The key is the synergy between humans and machines
What kind of synergy?
People are producers and customers
knowledge sources
have real world problems and interests
learn/create knowledge communicating with each other
Machines are enablers
store & remember data
search & combine data
draw mathematical & logical inferencex
The Vision ofCollective Intelligence
With the rise of the social web we have now millions of humans offering their knowledge online i.e.
The information is stored, searchable and easily shared
Challenge: match between what is put online and methods for doing useful reasoning with data
True collective knowledge emerges if the knowledge collected from all those people is aggregated or recombined to create new knowledge or new ways of learning
Collective Knowledge Systems
human-machines systems in which machines enable the collection and harvesting of large amounts of human-generated knowledge
Collective knowledge systemsthe faq-o-sphere
social system supported by ICT which generates self-service problem solving discussions in the internet
product support forums
special interest mailing lists
structured question-answer catalogs
in which some people pose problems and others reply with answers
Collective Knowledge Systems the faq-o-sphere
A search engine able of finding questions and answers in this body of content
Google is very good in finding a message in public forums in which someone has asked a question similar to one’s query
intelligent users, who know how to formulate their queries and provide feedback about which query/doc pairs were effective
though not designed as a system, faq-o-sphere behave as competent expert systems
Collective Knowledge Systemsthe faq-o-sphere
Collective Knowledge Systems
Citizen Journalism
blog-o-sphere
Product Reviews
computer products, gadgets, digital cameras
Collaborative filtering
Amazon recomendations
Collective Knowledge Systems
User-generated content (by a lot of users!)
Human-machine synergy
Increasing returns with scale
Emergent Knowledge
new ideas, products, concepts, theories, ways of doing things, etc.
how? with the semantic web
Semantic Web
The problem of semantics
what we say
how we say it
different symbols/terms with same meaning
same symbols/terms with different meaning
Traditional web
htmlkeyword-based searh
ProblemComputers don’t understand meaning
Solution?
“My mouse is broken. I need a new one…”
Semantic Web
Semantic Web Benefits
Semantic Web Layer Cake
The Web of Things
RDF
URIs
RDF -> XML
Ontologies
Concept conceptual entity of the domain
Attribute property of a concept
Relation relationship between concepts or properties
Axiom coherent description between Concepts / Properties / Relations via logical expressions
Person
Student Professor
Lecture
isA – hierarchy (taxonomy)
name email
student nr.
research field
topic lecture nr.
attends holds
The role of the semantic webTechnology has enabled the generation of collected knowledge by making it easy and cheap to:
Capture
Store
Distribute
Communicate
Create new value from the collected data
The role of the semantic webCreating value from data is the main role of the semantic web in collective knowledge systems
semantic web adds structure to data related to user contributions
enabling sharing and computation among independent, heterogeneous social web applications
The role of the semantic web
Augmenting user-contributed data with structured data
structured data exposed in a structured way
distinguish Paris Hilton from Paris, France
expose data in data bases used to build html documents
extract data retrospectively from user contributions
capture data as people share information
Enabling data sharing and computation among applications
RDF enables structured data referencing well maintained namespaces, unambiguous entity reference with URIs
Ontologies for common conceptualizations independent of data models
in social web applications enables integrating tagging data
tagCommons project (mapping rather than homogenizing)
The role of the semantic web
Example: Real Travel
RealTravel attracts people to write about their travels, sharing stories, photos, etc.
Travel researchers get the value of all experiences relevant to their target destinations.
Real Travel
Real Travel
Real Travel
Real Travel
Group Stories together by destination
Aggregate cities to states to countries
Inherit locatioins down to photos
Infer geo-coordinatees, which drive dynamic rout management
Destinations map
to external contents (travel guides)
to targeted advertising
Real Travel as Collective Knowledge System
User generated content
Most of the content is from real traveler experiences
Human-machine synergy
travel planners could do the equivalent asking asking thousands of other travelers advice
Increasing returns with scale
as more people report their experiences, better coverage (more exotic locations) and depth (what to do or avoid)
Emergent knowledge
recommendations from unsupervised learning from travel blog texts and multi-dimensional match with structured data (e.g. traveler demographics, declared interest)
Real Traveler as Collective Knowledge Systems
Snap to grid Travel Destinations
auto-completion of candidate locations
allow introducing new locations
Contextual browsing
combining tags, location and rating data (feedback from users and editors of content quality)
Snap to grid Tags
associate tags to useful domain concepts (e.g. arts)
Real Travel Collective Knowledge System
Real TravelPivot searching
Structured data provides dimensions of a hypercube
location, author, type, date, quality rating
Travel researchers browse along any dimension.
The key structured data is the destination hierarchy
Contributors place their content into the destination hierarchy, and the other dimensions are automatic.
Real Travel as Collective Knowledge Systems
Learning from semi-structured data
System processes every contribution looking at text, tags, user profiles and other structured data
Clustering of the content to find synthetic dimensions
Stable classification of blogs and users in buckets
when users ask for recommendations they introduce desired location, trip length and demographic data
this data is used to filter some dimensions and they are asked to rate the remaining dimensions
the system matches this information with classified users and docs and ranks places to go and traveler blogs for those places
Resources used
Open source software or free services
powerful databases
fancy UI libraries
search engines
usage analytics
Open APIs from Google Maps and Flickr (photos)
Commercially available geo-coordinate data and services
How could semantic could help?No standard source of structured destination data for the world
or way to map among alternative hierarchies
Integrating with other destination-based sites is expensive
e.g. travel guides
No standard collection of travel tags
or way to share RealTravel’s folksonomy
Integration with other tagging sites is ad-hoc
top related