web characteristics - Professor C. Lee Giles

advertisement
Web characteristics, broadly
defined
Thanks to C. Manning, P. Raghavan, H.
Schutze
What we have covered









What is IR
Evaluation
Tokenization and properties of text
Web crawling
Vector methods
Measures of similarity
Indexing
Inverted files
Today:

Web characteristics






Web vs classic IR
Web advertising
Web as a graph – size
SEO
Web spam
Dup detection
Query Engine
Index
Interface
Indexer
Users
Crawler
Characteristics
Web
A Typical Web Search Engine
What Is the World Wide Web?

The world wide web (web) is a network of information resources,
most of which is published by humans.
–

World’s largest publishing mechanism
The web relies on three mechanisms to make these resources
readily available to the widest possible audience:
–
A uniform naming scheme for locating resources on the web (e.g.,
URIs).
•
–
Protocols, for access to named resources over the web (e.g., HTTP).
•
–
How to get there
Hypertext, for easy navigation among resources (e.g., HTML).
•

Find it
Where to go from there.
Web is opt-out not an opt-in system
–
–
Your information is available to all; you have to protect it.
DMCA (Digital Millennium Copyright Act)
–
Only to those who sign it.
Internet vs. Web

Internet:
•
Internet is a more general term
Includes physical aspect of underlying
networks and mechanisms such as email,
FTP, HTTP…
•

Web:
•
HTTP
Associated with information stored on the
Internet
Refers to a broader class of networks, i.e.
Web of English Literature
•
•

Both Internet and web are networks
Networks vs
Graphs
Examples?
Old internet network
Internet Technologies
Web Standards

Internet Engineering Task Force (IETF)




http://www.ietf.org/
Founded 1986
Request For Comments (RFC) at
http://www.ietf.org/rfc.html
World Wide Web Consortium (W3C)



http://www.w3.org
Founded 1994 by Tim Berners-Lee
Publishes technical reports and recommendations
Internet Technologies
Web Design Principles
Interoperability: Web languages and
protocols must be compatible with one
another independent of hardware and
software.
Evolution: The Web must be able to
accommodate future technologies.




Encourages simplicity, modularity and
extensibility.
Decentralization: Facilitates scalability and
robustness.
Languages of the WWW

Markup languages


A markup language combines text and extra
information about the text. The extra information,
for example about the text's structure or
presentation, is expressed using markup, which is
intermingled with the primary text.
The best-known markup language is in modern
use is HTML (Hypertext Markup Language), one
of the foundations of the World Wide Web.
Historically, markup was (and is) used in the
publishing industry in the communication of
printed work between authors, editors, and
printers.
Without search engines the web wouldn’t
scale (there would be no web)
1.
2.
3.
No incentive in creating content unless it can be easily found –
other finding methods haven’t kept pace (taxonomies,
bookmarks, etc)
The web is both a technology artifact and a social environment
 “The Web has become the “new normal” in the American way
of life; those who don’t go online constitute an ever-shrinking
minority.” – [Pew Foundation report, January 2005]
Search engines make aggregation of interest possible:
 Create incentives for very specialized niche players


4.
5.
Economical – specialized stores, providers, etc
Social – narrow interests, specialized communities, etc
The acceptance of search interaction makes “unlimited
selection” stores possible:

Amazon, Netflix, etc
Search turned out to be the best mechanism for advertising on
the web, a $15+ B industry (2011)
 Growing very fast but entire US advertising industry $250B –
huge room to grow
 Sponsored search marketing is about $10B
Classical IR vs. Web IR
Basic assumptions of
Classical Information Retrieval



Corpus: Fixed document collection
Goal: Retrieve documents with information
content that is relevant to user’s information
need
Searcher: information scientist or a search
professional trained in making logical queries
Classic IR Goal

Classic relevance



For each query Q and stored document D in a given
corpus assume there exists relevance Score(Q, D)
 Score is average over users U and context C
Optimize Score(Q, D) as opposed to Score(Q, D, U, C)
That is, usually:
 Context ignored
 Individuals ignored
Bad assumptions
in the web context
 Corpus predetermined
Web Information Retrieval
Basic assumptions of
Web Information Retrieval


Corpus: constantly changing; created by
amateurs and professionals
Goal: Retrieve summaries of relevant information
quickly with links to the original site


High precision! Recall not important
Searcher: amateurs; no professional training and
less or no concern about quality queries
Subscription
Feeds
Crawls
Content creators
Transaction
Advertisement
Editorial
The coarse-level dynamics
Content aggregators
Content consumers
Brief (non-technical) history

Early keyword-based engines


Altavista, Excite, Infoseek, Inktomi, ca. 1995-1997
Paid placement ranking: Goto.com (morphed into
Overture.com  Yahoo!)


Your search ranking depended on how much you
paid
Auction for keywords: casino was expensive!
Brief (non-technical) history

1998+: Link-based ranking pioneered by Google




Blew away all early engines save Inktomi
Great user experience in search of a business
model
Meanwhile Goto/Overture’s annual revenues
were nearing $1 billion
Result: Google added paid-placement “ads” to
the side, independent of search results

Yahoo follows suit, acquiring Overture (for paid
placement) and Inktomi (for search)
Query Results 2009
Ads
Algorithmic
results.
Query Results 2010
Ads
Algorithmic
results.
Query Results 2013
Algorithmic
results.
Query Results 2014
Algorithmic
results.
Algorithmic
results.
Query Results 2016
Algorithmic
results.
Query String for Google 2013
http://www.google.com/search?q=lee+giles
“?” in query implies the start of a query component
query equals lee AND giles
Basic query reduced from the full one.
Query above is
lee giles
What does Bing do?
duckduckgo?
• For more, see wikipedia query string
Ads vs. search results

Google has maintained that ads
(based on vendors bidding for
keywords) do not affect vendors’
rankings in search results
Web
Search =
miele
Sponsored Links
CG Appliance Express
Discount Appliances (650) 756 -3931
Same Day Certified Installation
www.cgappliance.com
San Francis co -Oakland - San Jose,
CA
Miele Vacuum Cleaners
Miele Vacuums - Complete Sele ction
Free Shipping!
www.vacuums.com
Miele Vacuum Cleaners
Miele -Free Air shipping!
All models. Helpful advice.
www.best -vacuum.com
Results 1 - 10 of about 7,310,000 for miele . ( 0.12 seconds)
Miele , Inc -- Anything else is a compromise
At the heart of your home, Appliances by
Miele . ... USA. to miele .com. Residential Appliances.
Vacuum Cleaners. Dishwashers. Cooking Appliances. Steam Oven. Coffee System
...
www. miele .com/ - 20k - Cached - Similar pages
Miele
Welcome to Miele , the home of the very best appliances and kitchens in the world.
www. miele .co.uk/ - 3k - Cached - Similar pages
Miele - Deutscher Hersteller von
Einbaugeräten, Hausgeräten ... - [ Translate this
page ]
Das Portal zum Thema Essen & Geniessen online unter www.zu
-tisch.d e. Miele weltweit
...ein Leben lang. ... Wählen Sie die Miele Vertretung Ihres Landes.
www. miele .de/ - 10k - Cached - Similar pages
Miele Österreich - [ Translate this page
Herzlich willkommen bei Miele Österreich Wenn Sie nicht automatisch
weitergeleitet werden, klicken Sie bitte hier! HAUSHALTSGERÄTE
...
www. miele .at/ - 3k - Cached - Similar pages
Herzlich willkommen bei
]

Google has maintained that ads (based on vendors
bidding for keywords) do not affect vendors’ rankings
in search results

Search 2016
Ads vs. search results

Other vendors (Yahoo, Bing) have made
similar statements from time to time


Any of them can change anytime
Focus primarily on search results
independent of paid placement ads

Although the latter is a fascinating
technical subject in itself
Pay Per Click (PPC) Search
Engine Ranking






PPC ads appear as “sponsored listings”
Companies bid on price they are willing to pay “per click”
Typically have very good tracking tools and statistics
Ability to control ad text
Can set budgets and spending limits
Google AdWords and Overture are the two leaders

Google $67B income
from advertising

Most expensive ad
words

Research field:


Computational
advertising
Algorithmic
advertising
Web search basics
Sponsored Links
CG Appliance Express
Discount Appliances (650) 756 -3931
Same Day Certified Installation
www.cgappliance.com
San Francisco -Oakland - San Jose,
CA
User
Miele Vacuum Cleaners
Miele Vacuums - Complete Selection
Free Shipping!
www.vacuums.com
Miele Vacuum Cleaners
Miele -Free Air shipping!
All models. Helpful advice.
www.b est -vacuum.com
Web
Results 1 - 10 of about 7,310,000 for miele . ( 0.12 seconds)
Miele, Inc -- Anything else is a compromise
Web spider
At the heart of your home, Appliances by
Miele . ... USA. to miele .com. Residential Appliances.
Vacuum C leaners. Dishwashers. Cooking Appliances. Steam Oven. Coffee System
...
www. miele .com/ - 20k - Cached - Similar pages
Miele
Welcome to Miele , the home of the very best appliances and kitchens in the world.
www. miele .co.uk/ - 3k - Cached - Similar pages
Miele - Deutscher Hersteller von Einbaugeräten, Hausgeräten
... - [ Translate this
page ]
Das Portal zum Thema Essen & Geniessen online unter www.zu
-tisch.de. Miele weltweit
...ein Leben lang. ... Wählen Sie die Miele Vertretung Ihres Landes.
www. miele .de/ - 10k - Cached - Similar pages
Miele Österreich - [ Translate this page ]
Herzlich willkommen bei Miele Österreich Wenn Sie nicht automatisch
weitergeleitet werden, klicken Sie bitte hier! HAUSHALTSGERÄTE
...
www. miele .at/ - 3k - Cached - Similar pages
Herzlich willkommen bei
Search
Indexer
The Web
Indexes
Ad indexes
User Needs

Need [Brod02, RL04]




Informational – want to learn about something (~40% / 65%)
Low hemoglobin
Navigational – want to go to that page (~25% / 15%)
United Airlines
Transactional – want to do something (web-mediated) (~35% / 20%)

Access a service

Downloads

Shop
Seattle weather
Mars surface images
Canon S410
Gray areas


Car rental Brasil
Find a good hub
Exploratory search “see what’s there”
Web search users

Make ill defined
queries

Short









Specific behavior

AV 2001: 2.54 terms
avg, 80% < 3 words)
AV 1998: 2.35 terms
avg, 88% < 3 words
[Silv98]
Imprecise terms
Sub-optimal syntax (most
queries without operator)
Low effort
Wide variance in


Needs
Expectations
Knowledge
Bandwidth


85% look over one
result screen only
(mostly above the
fold)
78% of queries are
not modified (one
query/session)
Follow links –
“the scent of
information” ...
Query Distribution
Power law: few popular broad queries,
many rare specific queries
How far do people look for results?
(Source: iprospect.com WhitePaper_2006_SearchEngineUserBehavior.pdf)
Users’ empirical evaluation of results

Quality of pages varies widely


Relevance is not enough
Other desirable qualities (non IR!!)




Precision vs. recall


On the web, recall seldom matters
What matters


Precision at 1? Precision above the fold?
Comprehensiveness – must be able to deal with obscure
queries


Content: Trustworthy, new info, non-duplicates, well maintained,
Web readability: display correctly & fast
No annoyances: pop-ups, etc
Recall matters when the number of matches is very small
User perceptions may be unscientific, but are
significant over a large aggregate
Users’ empirical evaluation of engines




Relevance and validity of results
UI – Simple, no clutter, error tolerant
Trust – Results are objective
Coverage of topics for polysemic queries
Eg. mole, bank

Pre/Post process tools provided




Mitigate user errors (auto spell check, syntax errors,…)
Explicit: Search within results, more like this, refine ...
Anticipative: related searches
Deal with idiosyncrasies

Web specific vocabulary



Impact on stemming, spell-check, etc
Web addresses typed in the search box
…
The Web corpus
The Web
Large dynamic directed graph
 No design/co-ordination
 Distributed content creation, linking,
democratization of publishing
 Content includes truth, lies, obsolete
information, contradictions …
 Unstructured (text, html, …), semistructured (XML, annotated photos),
structured (Databases)…
 Scale much larger than previous text
corpora … but corporate records are
catching up.
 Growth – slowed down from initial
“volume doubling every few months”
but still expanding
 Content can be dynamically
generated
Web Today

The Web consists of hundreds of billions of
pages (Google query ‘the’)




Potentially infinite if dynamic pages are
considered
It is considered one of the biggest information
revolutions in recent human history
One of the largest graphs around
Full of information

Trends
Web page
The simple Web graph

A graph G = (V, E) is defined by



The Web page graph (directed)



a set V of vertices (nodes)
a set E of edges (links) = pairs of nodes
V is the set of static public pages
E is the set of static hyperlinks
Many more graphs can be defined




The host graph
The co-citation graph
Temporal graph
etc
Which pages do we care for if we
want to measure the web graph?

Avoid “dynamic” pages?




catalogs
pages generated by queries
pages generated by cgi-scripts (the nostradamus
effect)
Only interested in “static” web pages
The Web: Dynamic content

A page without a static html version



E.g., current status of flight AA129
Current availability of rooms at a hotel
Usually, assembled at the time of a request from a browser


Sometimes, URL has a ‘?’ character in it
‘?’ precedes the actual query
AA129
Browser
Application server
Back-end
databases
The Static Public Web
Example - http:/clgiles.ist.psu.edu
Static

Public
no password
required
 no robots.txt

exclusion

 no “noindex” meta
tag

 These rules can still be fooled etc.
 “Dynamic pages” appear static

not the result of a cgibin scripts
no “?” in the URL
doesn’t change very
often
etc.


• browseable catalogs (Hierarchy built from DB)
 Spider traps -- infinite url descent
• www.x.com/home/home/home/…./home/home.html
 Spammer games
Why do we care about the Web graph?


Is it the largest human artifact ever created?
Exploit the Web structure for







crawlers
search and link analysis ranking
spam detection
community discovery
classification/organization
business, politics, society applications
Predict the Web future





mathematical models
algorithm analysis
sociological understanding
New business opportunities
New politics
The first question: what is the size of the Web?

Surprisingly hard to answer

Naïve solution: keep crawling until the whole graph has been
explored

Extremely simple but wrong solution: crawling is complicated
because the web is complicated




spamming
duplicates
mirrors
Simple example of a complication: Soft 404


When a page does not exists, the server is supposed to return an
error code = “404”
Many servers do not return an error code, but keep the visitor on
site, or simply send him to the home page
A sampling approach

Sample pages uniformly at random
Compute the percentage of the pages that
belong to a search engine repository (search
engine coverage)
Estimate the size of the Web

Problems:




how do you sample a page uniformly at random?
how do you test if a page is indexed by a search
engine?
Sampling pages [LG99, HHMN00]

Create IP addresses uniformly at random


problems with virtual hosting, spamming
Starting from a subset of pages perform a
random walk on the graph. After “enough” steps
you should end up in a random page.

near uniform sampling
Testing search engine containment [BB98]
Measuring the Web

It is clear that the Web that we see is what
the crawler discovers

We need large crawls in order to make
meaningful measurements
The measurements are still biased by




the crawling policy
size limitations of the crawl
Perturbations of the "natural" process of birth and
death of nodes and links
Measures on the Web graph [BKMRRSTW00]


Degree distributions
The global picture





what does the Web look like from a far?
Reachability
Connected components
Community structure
The finer picture
In-degree distribution

Power-law distribution with exponent 2.1
Out-degree distribution

Power-law distribution with exponent 2.7
The good news

The fact that the exponent is greater than 2
implies that the expected value of the degree is a
constant (not growing with n)

Therefore, the expected number of edges is
linear in the number of nodes n

This is good news, since we cannot handle
anything much more than linear
Connected components – definitions

Weakly connected components (WCC)


Set of nodes such that from any node can go to any node via an
undirected path
Strongly connected components (SCC)

Set of nodes such that from any node can go to any node via a
directed path.
WCC
SCC
The bow-tie structure of the Web
SCC and WCC distribution

The SCC and WCC sizes follows a power law
distribution

the second largest SCC is significantly smaller
The inner structure of the bow-tie

[LMST05]
What do the individual components of the bow tie
look like?

They obey the same power laws in the degree
distributions
The inner structure of the bow-tie

Is it the case that the bow-tie repeats itself in
each of the components (self-similarity)?


It would look nice, but this does not seem to be the
case
no large WCC, many small ones
The daisy structure?

Large connected core, and highly fragmented
IN and OUT components

Unfortunately, we do not have a large crawl
to verify this hypothesis
A different kind of self-similarity [DKCRST01]

Consider Thematically Unified Clusters (TUC):
pages grouped by






keyword searches
web location (intranets)
geography
hostgraph
random collections
All such TUCs exhibit a bow-tie structure!
Self-similarity

The Web consists of a collection of self-similar
structures that form a backbone of the SCC
Dynamic content

Most dynamic content is ignored by web spiders


Some dynamic content (news stories from
subscriptions) are sometimes delivered as dynamic
content



Many reasons including malicious spider traps
Application-specific spidering
Spiders commonly view web pages just as Lynx (a text
browser) would
Note: even “static” pages are typically assembled on
the fly (e.g., headers are common)
The web: size

What is being measured?


Number of hosts
Number of (static) html pages


Number of hosts – netcraft survey



Volume of data
http://news.netcraft.com/archives/web_server_survey.html
Monthly report on how many web hosts & servers are out
there
Number of pages – numerous estimates
Netcraft Web Server Survey
http://news.netcraft.com/archives/web_server_survey.html
Netcraft Web Server Survey
http://news.netcraft.com/archives/web_server_survey.html
Netcraft Web Server Survey
http://news.netcraft.com/archives/web_server_survey.html
Netcraft Web Server Survey
http://news.netcraft.com/archives/web_server_survey.html
The web: evolution


All of these numbers keep changing
Relatively few scientific studies of the
evolution of the web [Fetterly & al, 2003]


http://research.microsoft.com/research/sv/svpubs/p97-fetterly/p97-fetterly.pdf
Sometimes possible to extrapolate from
small samples (fractal models) [Dill & al,
2001]

http://www.vldb.org/conf/2001/P069.pdf
Rate of change

[Cho00] 720K pages from 270 popular sites sampled
daily from Feb 17 – Jun 14, 1999


[Fett02] Massive study 151M pages checked over
few months



Any changes: 40% weekly, 23% daily
Significant changed -- 7% weekly
Small changes – 25% weekly
[Ntul04] 154 large sites re-crawled from scratch
weekly




8% new pages/week
8% die
5% new content
25% new links/week
Static pages: rate of change

Fetterly et al. study (2002): several views of data, 150 million
pages over 11 weekly crawls

Bucketed into 85 groups by extent of change
Other characteristics

Significant duplication



High linkage


More than 8 links/page in the average
Complex graph topology


Syntactic – 30%-40% (near) duplicates [Brod97,
Shiv99b, etc.]
Semantic – ???
Not a small world; bow-tie structure [Brod00]
Spam

Billions of pages
Web
evolution
Spam
vs
Search Engine Optimization (SEO)
SERPs
• Search engine results
page (SERP) are
very influential
• How can they be
manipulated
• How can you come
up first?
Google SERP for
query “lee giles’
The trouble with paid placement…


It costs money. What’s the alternative?
Search Engine Optimization (SEO):






“Tuning” your web page to rank highly in the
search results for select keywords
Alternative to paying for placement
Thus, intrinsically a marketing function
Performed by companies, webmasters and
consultants (“Search engine optimizers”) for
their clients
Some perfectly legitimate, some very shady
Some frowned upon by Search Engines
Simplest forms

First generation engines relied heavily on tf/idf


The top-ranked pages for the query maui resort were the
ones containing the most maui’s and resort’s
SEOs responded with dense repetitions of chosen
terms


e.g., maui resort maui resort maui resort
Often, the repetitions would be in the same color as the
background of the web page


Repeated terms got indexed by crawlers
But not visible to humans on browsers
Pure word density cannot
be trusted as an IR signal
Search Engine Spam: Objective
Success of commercial Web sites depends on the number of
visitors that find the site while searching for a particular product.
85% of searchers look at only the first page of results
A new business sector – search engine optimization
M. Henzinger, R. Motwani, and C. Silverstein. Challenges in web
search engines. International Joint Conference on Artificial
Intelligence, 2003.
Drost, I. and Scheffer, T., Thwarting the Nigritude Ultramarine:
Learning to Identify Link Spam. 16th European Conference on
Machine Learning, Porto, 2005
What’s SEO?

SEO = Search Engine Optimization


Refers to the process of “optimizing” both the on-page and offpage ranking factors in order to achieve high search engine
rankings for targeted search terms.
Refers to the “industry” that revolves around obtaining high
rankings in the search engines for desirable keyword search terms
as a means of increasing the relevant traffic to a given website.

Refers to an individual or company that optimizes websites for its
clientele.

Has a number of related meanings, and usually refers to an
individual/firm that focuses on optimizing for “organic” search
engine rankings
What’s SEO based on

Features we know used in web page
ranking






Page content
Page metadata
Anchor text
Links
User behavior
Others?
Search Engine Spam: Spamdexing
spamdexing (also known as search spam, search
engine spam, web spam or search engine
poisoning) is the deliberate manipulation of search
engine indexes
aka: black hat SEO
Porn led the way – ‘96
Search Engine Spamdexing Methods
Content based
Link spam
Cloaking
Mirror sites
URL redirection
Content spamming
Keyword stuffing
•calculated placement of keywords within a page to raise the keyword count,
variety, and density of the page. Truncation so that massive dictionary lists
cannot be indexed on a single webpage.
•Hidden or invisible text
Meta-tag stuffing
•Out of date
Doorway pages
•"Gateway" or doorway pages created with very little content but are instead
stuffed with very similar keywords and phrases.
•BMW caught
Scraper sites
•amalgamation of content taken from other sites; still works
Article spinning
•rewriting existing articles, as opposed to merely scraping content from other
sites, undertaken by hired writers or automated using a thesaurus database or a
neural network.
Variants of keyword stuffing


Misleading meta-tags, excessive repetition
Hidden text with colors, style sheet tricks,
etc.
Meta-Tags =
“… London hotels, hotel, holiday inn, hilton, discount,
booking, reservation, sex, mp3, britney spears, viagra,
…”
Link Spamming: Techniques
Link farms: Densely connected arrays of pages. Farm pages
propagate their PageRank to the target, e.g., by a funnelshaped architecture that points directly or indirectly towards
the target page. To camouflage link farms, tools fill in
inconspicuous content, e.g., by copying news bulletins.
Link exchange services: Listings of (often unrelated)
hyperlinks. To be listed, businesses have to provide a back
link that enhances the PageRank of the exchange service.
Guestbooks, discussion boards, and weblogs: Automatic
tools post large numbers of messages to many sites; each
message contains a hyperlink to the target website.
Cloaking


Serve fake content to search engine spider
DNS cloaking: Switch IP address. Impersonate
Y
SPAM
Is this a Search
Engine spider?
Cloaking
N
Real
Doc
Google Bombing != Google Hacking



http://en.wikipedia.org/wiki/Google_bomb
A Google bomb or Google wash is an attempt to
influence the ranking of a given site in results
returned by the Google search engine. Due to the
way that Google's Page Rank algorithm works, a
website will be ranked higher if the sites that link to
that page all use consistent anchor text.
A google bomb is when links are put in several sites
on the internet by a number of people so it leads a
particular keyword search combination to a specific
site and the site is deluged/swamped (it may crash
the site) with hits.
94
Google Bomb - old
Query: french military victories
Others?
Link Spamming: Defenses
Manual identification of spam pages and farms to create a
blacklist.
Automatic classification of pages using machine learning
techniques.
BadRank algorithm. The "bad rank" is initialized to a high
value for blacklisted pages. It propagates bad rank to all
referring pages (with a damping factor) thus penalizing
pages that refer to spam.
Intelligent SEO


Figure out how search engines do their ranking
Inductive science



Make intelligent changes
Figure our what happens
Repeat
So What is the Search Engine
Ranking Algorithm?





Top Secret! Only select employees of the actual search
engines know for certain
Reverse engineering, research and experiments gives
some idea of major factors and approximate weight
assignments
Constant changing, tweaking, updating is done to the
algorithm
Websites and documents being searched are also
constantly changing
Varies by Search Engine – some give more weight to onpage factors, some to link popularity
SEO Expert
Search engine optimization vs Spam

Motives



Operators




Commercial, political, religious, lobbies
Promotion funded by advertising budget
Contractors (Search Engine Optimizers) for lobbies, companies
Web masters
Hosting services
Forums

E.g., Web master world ( www.webmasterworld.com )


Search engine specific tricks
Discussions about academic papers 
The spam industry
SEO contests
•Now part of some search engine classes!
The war against spam

Quality signals - Prefer
authoritative pages based
on:








Ignore statistically implausible
linkage (or text)
Use link analysis to detect
spammers (guilt by
association)

Training set based on
known spam
Family friendly filters

Anti robot test
Limits on meta-keywords
Robust link analysis
Spam recognition by
machine learning

Policing of URL
submissions


Votes from authors (linkage
signals)
Votes from users (usage
signals)

Linguistic analysis, general
classification techniques,
etc.
For images: flesh tone
detectors, source text
analysis, etc.
Editorial intervention




Blacklists
Top queries audited
Complaints addressed
Suspect pattern detection
The war against spam – what Google does

Google changes their algorithms regularly

Panda and Penguin updates



Suing Google doesn’t work


Google Wins in SearchKing Lawsuit
Defines what counts as a high quality site


Uses machine learning and AI methods
SEO doesn’t seem to use this
23 features
Google penalty



refers to a negative impact on a website's search ranking
based on updates to Google's search algorithms.
penalty can be an unfortunate by-product
an intentional penalization of various black-hat SEO
techniques
Future of spamdexing

Who’s the smartest tech wise: Search engines
vs SEOs



Web search engines have policies on SEO
practices they tolerate/block



http://www.bing.com/toolbox/webmaster
http://www.google.com/intl/en/webmasters/
Adversarial IR: the unending (technical) battle
between SEO’s and web search engines


Constant evolution
Can it ever be solved?
Research http://airweb.cse.lehigh.edu/
Many SEO companies will suffer
Reported “Organic” (white hat)
Optimization Techniques












Register with search engines
Research keywords related to your business
Identify competitors, utilize benchmarking techniques and identify level of
competition
Utilize descriptive title tags for each page
Ensure that your text is HTML text and not image text
Use text links when possible
Use appropriate keywords in your content and internal hyperlinks (don’t
overdo!)
Obtain inbound links from related websites
Use Sitemaps
Monitor your search engine rankings and more importantly your website
traffic statistics and sales/leads produced
Educate yourself about search engine marketing or consult a search
engine optimization firm or SEO expert
Use the Google Guide to High Quality Websites
Duplicate detection
Get rid of duplicates; save space and time
Product or idea evolution (near duplicates)
Check for stolen information, plagiarism, etc.

Duplicate documents

The web is full of duplicated content


Strict duplicate detection = exact match


Estimates at 40%
Not as common
But many, many cases of near duplicates

E.g., Last modified date the only difference
between two copies of a page
Duplicate/Near-Duplicate Detection

Duplication: Exact match can be detected with
fingerprints


Hash value
Near-Duplication: Approximate match

Overview


Compute syntactic similarity with an edit-distance
measure
Use similarity threshold to detect near-duplicates


E.g., Similarity > 80% => Documents are “near
duplicates”
Not transitive though sometimes used transitively
Computing Similarity


Features:
 Segments of a document (natural or artificial breakpoints)
 Shingles (Word N-Grams)
 a rose is a rose is a rose →
a_rose_is_a
rose_is_a_rose
is_a_rose_is
a_rose_is_a
Similarity Measure between two docs (= sets of shingles)
 Set intersection
 Specifically (Size_of_Intersection / Size_of_Union)
Shingles + Set Intersection
Computing exact set intersection of shingles
between all pairs of documents is
expensive/intractable


Approximate using a cleverly chosen subset of
shingles from each (a sketch)
Estimate (size_of_intersection / size_of_union)
based on a short sketch

Doc
A
Shingle set A
Doc
B
Shingle set A
Sketch A
Jaccard
Sketch A
Sketch of a document

Create a “sketch vector” (of size ~200) for
each document


Documents that share ≥ t (say 80%)
corresponding vector elements are near
duplicates
For doc D, sketchD[ i ] is as follows:



Let f map all shingles in the universe to 0..2m
(e.g., f = fingerprinting)
Let pi be a random permutation on 0..2m
Pick MIN {pi(f(s))} over all shingles s in D
Comparing Signatures


Signature Matrix S
 Rows = Hash Functions
 Columns = Columns
 Entries = Signatures
Can compute – Pair-wise similarity of
any pair of signature columns
All signature pairs


Now we have an extremely efficient method for
estimating a Jaccard coefficient for a single pair
of documents.
But we still have to estimate N2 coefficients
where N is the number of web pages.



Still slow
One solution: locality sensitive hashing (LSH)
Another solution: sorting (Henzinger 2006)
What we covered




Web vs classic IR
What users do
Web advertising
Web as a graph



SEO/Web spam


Applications
Structure and size
White vs black hat
Dup detection
Download