Making the Web searchable, or the Future of Web Search

1
Making the Web searchable, or the Future of Web
Search

• Peter Mika
• Yahoo! Research Barcelona

2
About Yahoo!

• Yahoo!'s mission is to connect people to their
  passions, their communities, and the worlds
  knowledge
• Yahoo! Research Barcelona
• Established January, 2006
• Led by Ricardo Baeza-Yates
• Research areas
• Web Mining
• content, structure, usage
• Distributed Web retrieval
• Multimedia retrieval
• NLP and Semantics

3
Yahoo! by numbers (April, 2007)

• There are approximately 500 million users of
  Yahoo! branded services, meaning we reach 50
  percent or 1 out of every 2 users online, the
  largest audience on the Internet (Yahoo! Internal
  Data).
• Yahoo! is the most visited site online with
  nearly 4 billion visits and an average of 30
  visits per user per month in the U.S. and leads
  all competitors in audience reach, frequency and
  engagement (comScore Media Metrix, US, Feb.
  2007).
• Yahoo! accounts for the largest share of time
  Americans spend on the Internet with 12 percent
  (comScore Media Metrix, US, Feb. 2007) and
  approximately 8 percent of the worlds online
  time (comScore WorldMetrix, Feb. 2007).
• Yahoo! is the 1 home page with 85 million
  average daily visitors on Yahoo! homepages around
  the world, an increase of nearly 5 million
  visitors in a month (comScore WorldMetrix, Feb.
  2007).
• Yahoo!s social media properties (Flickr,
  delicious, Answers, 360, Video, MyBlogLog,
  Jumpcut and Bix) have 115 million unique visitors
  worldwide (comScore WorldMetrix, Feb. 2007).
• Yahoo! Answers is the largest collection of human
  knowledge on the Web with more than 90 million
  unique users and 250 million answers worldwide
  (Yahoo! Internal Data).
• There are more than 450 million photos in Flickr
  in total and 1 million photos are uploaded daily.
  80 percent of the photos are public (Yahoo!
  Internal Data).
• Del.icio.us hits 2 million users in February,
  growing more than six times its size from 300,000
  users in December 2005 (Yahoo! Internal Data).
• Yahoo! Mail is the 1 Web mail provider in the
  world with 243 million users (comScore
  WorldMetrix, Feb. 2007) and nearly 80 million
  users in the U.S. (comScore Media Metrix, US,
  Feb. 2007)
• Interoperability between Yahoo! Messenger and
  Windows Live Messenger has formed the largest IM
  community approaching 350 million user accounts
  (Yahoo! Internal Data).
• Yahoo! Messenger is the most popular in time
  spent with an average of 50 minutes per user, per
  day (comScore WorldMetrix, Feb. 2007).
• Nearly 1 in 10 Internet users is a member of a
  Yahoo! Groups (Yahoo! Internal Data).
• Yahoo! News is the 1 online news destination and
  has reached a new audience high in February with
  36.2 million users, 10 million more users than
  its nearest competitor MSNBC (comScore Media
  Metrix, US, Feb. 2007).
• Yahoo! is one of only 26 companies to be on both
  the Fortune 500 list and the Fortunes Best
  Place to Work List (2006).

4
Overview

• Why reconsider search?
• Context
• Semantic Web metadata infrastructure
• Web 2.0 user-generated metadata
• Thesis making the Web searchable
• Research challenges (SW IR)
• Conclusion

5
Motivation

• State of Web search
• Picked the low hanging fruit
• Heavy investments, marginal returns
• High hanging fruits
• Hard searches remain
• The Web and its technology have changed
• Semantic Web
• Web 2.0

6
Hard searches

• Ambiguous searches
• Paris Hilton
• Multimedia search
• Images of Paris Hilton
• Imprecise or overly precise searches
• Publications by Jim Hendler
• Find images of strong and adventurous people
  (Lenat)
• Searches for descriptions
• Search for yourself without using your name
• Product search (ads!)
• Searches that require aggregation
• Size of the Eiffer tower (Lenat)
• Public opinion on Britney Spears
• Queries that require a deeper understanding of
  the query, the content and/or the world at large
• Note some of these are so hard that users dont
  even try them any more

7
Example
8
The Semantic Web (1996-)

• Making the content of the Web machine processable
  through metadata
• Documents, databases, Web services
• Active research, standardization, startups
• Ontology languages (RDF, OWL family), query
  language for RDF (SPARQL)
• Software support (metadata stores, reasoners,
  APIs)

9
Problem difficulties in deployment

• Not enough take-up in the Web community at large
• Technological challenges
• Discovery
• Ontology learning
• Ontology mapping
• Lack of attention to the social side
• Over-estimating complexity for users
• Need for supporting ontology creation and sharing
• Focus shifts from documents to databases --the
  Web of Data
• Task/domain-specific applications

10
Web 2.0 (2003-)

• Simple, nimble, socially transparent interfaces
• Simplified KR
• e.g. tagging, microformats, Wikipedia infoboxes
• In exchange for a better experience,
• users are willing to
• Provide content, markup and metadata
• Provide data on themselves and their networks
• Rank, rate, filter, forward
• Develop software and improve your site
• User-generated content
• Content that users actually care about!

11
Example Microformats

• Agreements on the way to encode certain kinds
  metadata in HTML
• Reuse of semantic-bearing HTML elements
• Based on existing standards
• Each microformat defines a vocabulary for
  describing a given type of resource
• Persons, Events, but also syntactic metadata
  licenses, tags
• Not ontologies
• No formal descriptions of schema, only text
• No namespaces, unique identifiers (URIs)
• ? no interlinking, reuse among schemas
• No datatypes
• Widely used in millions of hand-authored
  documents
• And in hundreds of millions dynamically generated
  ones

12
Example hCard
href"mailtojfriday_at_host.com"Joe Friday
1-919-555-7878 class"title"Area Administrator, Assistant

rel"friend colleague met" href"http//meyerweb.
com/"Eric Meyer wrote a post
( /2005/12/16/tax-relief/" Tax Relief)
about an unintentionally humorous letter he
received from the class"fn org url" href"http//irs.gov/" Interna
l Revenue Service .
13
Example Wikipedia infoboxes

• Templates for common types of objects

14
Example Wikipedia infoboxes
15
Example Wikipedia infoboxes

• cf. microformats
• Similar level of representation
• Infoboxes are never annotations
• Largely uncontrolled growth
• Niche templates
• Templates in several languages
• ? overlapping domains
• Infoboxes to RDF
• dbPedia
• Compare also to Semantic Wikis
• Semantic MediaWiki, OntoWiki etc.

16
Web 2.0 bottleneck lack of foundations

• Tags
• No shared syntax (TagCommons? A microformat for
  tags?)
• Mapping problems due to lack of semantics
• flickrajax del.icio.usajax ?
• flickrajaxPeter flickrajaxJohn ?
• flickrajaxPeter1990 flickrajaxPeter2006 ?
• Microformats
• You cannot make a vocabulary for everything in a
  centralized way
• Serious validation,mapping problems on the
  instance level
• Wikipedia
• Serious validation,mapping problems on both the
  instance and the schema level

17
Thesis making the Web searchable

• The Web has changed
• Content owners are interested in their content to
  be found (Web 2.0)
• Cf. findability (Peter Morville), reusability
  (mashups), open data movement
• Foundations are laid for a Semantic Web
• We need to
• Combine the best of Web 2.0 and the Semantic Web
• Reconsider Web search in this new world

18
Semantic Web and Web 2.0

• Focus on user-generated content
• Getting the representation right
• RDF
• Embedded RDF
• GRDDL, RDFa
• Innovations on the interface side
• Capture semantics while authoring
• New methods of reasoning
• Semantics syntax statistics
• Bottom-up, emergent semantics
• Methods of logical reasoning combined with
  methods of graph mining, statistics
• Scalability
• Giving up soundness and/or completeness
• Dealing with the mess
• Social engineering
• Collaborative spaces for creating and sharing
  ontologies, data
• Connecting islands of semantics
• Best practices, documentation, advocacy

19
Example GRDDL

• Bridges the world of microformats and RDF
• Associate RDF-producing XSLT transformations to
  XML and (X)HTML documents
• One page may contain different microformats (e.g.
  persons and events described in the same page)
• One microformat may be mapped to multiple
  ontologies

ta-view" Joe Friday's Home page
href"http//www.w3.org/2003/12/rdf-in-xhtml-xslts
/grokFOAF.xsl" /

• Note of course it is possible to extract
  non-RDF data through XSLT,
• e.g. extract VCard from an HTML fragment ---
  but thats not called GRRDL

20
Example RDFa

• Embedding RDF into (X)HTML
• Increased complexity, e.g. namespaces
• Reuse of semantic-bearing HTML elements is not
  possible any more
• No need for XSLT any more
• You can use XSLT to extract RDFa, but dont have
  to
• Not much track record
• Big question user complexity (? data quality)

ons.bib" about"mika06jws"    
class"swrcArticle hbib article"    class"vcard"             property"foafname"       
rev"swrcauthor" href"mika06jws"       
class"foafPerson author fn" Peter Mika
       

21
Example openacademia.org and RDFa

• _at_INBOOKens03ontoknowledge,
• AUTHOR "Victor Iosif and Peter Mika and Rikard
  Larsson and Hans Akkermans",
• TITLE "Towards the Semantic Web
  Ontology-Driven Knowledge Management",
• CHAPTER "Field Experimenting With Semantic Web
  Tools In A Virtual Organization",
• PUBLISHER "John Wiley \ Sons",
• YEAR "2003

22
Example machine tags
23
Example ZoneTag project (Yahoo! Research
Berkeley)
24
Example Freebase
25
Web Search 2.0

• In an ideal world
• Plenty of metadata to harvest
• Metadata is unambiguous, described using a single
  ontology or a set of carefully designed
  ontologies
• User intent can be captured directly as a formal
  query
• Query and the knowledge base use the same
  ontology
• Query is executed on a single knowledge base,
  gives the correct, single answer
• And all this very fast

26
Web Search 2.0

• In reality
• Many lightweight ontologies or just tags
• Tags are mostly personal, not social
• Intent is unclear, matching is a problem
• Poor quality of annotations
• Everyones a librarian
• 99 of the Web is Web 1.0
• Input/output interface
• Keywords for searching
• Very limited interaction
• Not everything scales

27
Web Search 2.0

• Keep on improving machine technology
• NLP
• Information Extraction
• Exploit the users for the tasks that are hard for
  the machine
• Encourage and support users
• Exploit user-generated metadata in any shape or
  form
• Support standards of the SW architecture

28
Example folksonomies

• Simplified view tags are just anchortext
• Can be used to generate simple co-occurrence
  graphs

hilton
url1
paris
url2
eiffel
url3
29
(No Transcript)
30
The more complete picture

• Folksonomies as tripartite graphs of users, urls
  and tags

user1
user2
hilton
url1
user3
paris
url2
eiffel
url3
31
Mining and modelling folksonomies

• Opportunities for mining community-specific
  interpretations of the world
• Peter Mika. Ontologies are us A unified model of
  social networks and semantics. Journal of Web
  Semantics 5 (1), page 5-15, 2007
• Related works at
• Social and Collaborative Construction of
  Structured Knowledge (CKC2007)
• Bridging the gap between Semantic Web and Web 2.0
  (ESWC2007)
• Journal of Web Semantics special issue on
  Semantic Web and Web 2.0 (upcoming, Q4 2007)
• TAGORA project

32
Vision ontology-based search

• Query at the knowledge level
• Partial description of a class/instance
• Mapping of queries and resources in the
  conceptual space
• Computing relevance in semantic terms
• Novel user interfaces

33
Technical challenges

• Improving NLP and IE
• Query interface
• Data quality
• Cleaning up metadata, tags
• Spam
• Ontology mapping and entity resolution
• Ranking across types
• Results display
• How do you avoid information overload?
• How do you display information you partially
  understand?

34
Social challenges

• Getting the users on your side
• Users are unwilling to submit large amounts of
  structured data to a commercial entity (Google
  Base)
• Provide a clear motivation and/or instant
  gratification
• Trust them but not too much (Mahalo)

35
Example Technorati and microformats
http//technorati.com/posts/tag/semanticweb

• rel"tag"Semantic Web

36
Conclusion

• Why a new vision?
• The opportunity convergence
• Semantic Web metadata infrastructure
• Web 2.0 user-generated metadata
• Thesis making the Web searchable
• Semantic Web and Web 2.0
• Web Search 2.0

37
What is there to gain?

• Knowledge-based search
• Sorting out hard searches
• Creating new information needs
• Beyond search
• Analysis, design, diagnosis etc. on top of
  aggregated data
• Personalization
• Rich user profiles
• Monetization
• No more buy virgins on eBay

38
Questions?

• Peter Mika. Social Networks and the Semantic Web.
  Springer, July, 2007.
• Special Issue on the Semantic Web and Web 2.0,
  Journal of Web Semantics, December, 2007.