1

Web SearchFrom Information Retrieval to Microeconomic Modeling

Prabhakar Raghavan

Yahoo! Research

2Yahoo! Research

What is web search?

• Access to “heterogeneous”, distributed information– Heterogeneous in creation– Heterogeneous in accuracy– Heterogeneous in motives

• Multi-billion dollar business– Source of new opportunities in marketing

• Strains the boundaries of trademark and intellectual property laws

• A source of unending technical challenges

3Yahoo! Research

The coarse-level dynamics

Content creators Content aggregators

Feeds

Crawls

Content consumers

Adv

erti

sem

ent

Edi

tori

al

Sub

scri

ptio

nT

rans

acti

on

4Yahoo! Research

5Yahoo! Research

Brief (non-technical) history

• Early keyword-based engines

– Altavista, Excite, Infoseek, Inktomi, Lycos, ca. 1995-1997

• Paid placement ranking: Goto (morphed into Overture Yahoo!)

– Your search ranking depended on how much you paid

– Auction for keywords: casino was expensive!

6Yahoo! Research

Brief (non-technical) history

• 1998+: Link-based ranking pioneered by Google

– Blew away all early engines except Inktomi

– Great user experience in search of a business model

– Meanwhile Goto/Overture’s annual revenues were nearing $1 billion

7Yahoo! Research

Brief (non-technical) history

• Result: Google added “paid-placement” ads to the side, separate from search results

• 2003: Yahoo follows suit, acquiring Overture (for paid placement) and Inktomi (for search)

8Yahoo! Research

Algorithmic results CPC Advertisements

9Yahoo! Research

Editorial

User reviews

Ads

10Yahoo! Research

Types of content

• Editorial content: books, music, professionally-produced websites

• User-generated content: blogs, reviews, bulletin boards, groups, etc.

• Total web growth: 1-3M pages/day– Not “real” growth– Think text content…

Courtesy: Andrew Tomkins

11Yahoo! Research

Total text content

• 6B people type 4 hrs/day at 100 wpm

• Storage: 52PB/yr = Cost: $25M/yr

• In another 5 years, this looks about like the cost of having 10 people on your payroll

• Conclusion 1: any company with tens of people can store every bit of text produced by every human on the planet

• Conclusion 2: no scale-based differentiation around text content

(of course, not all content is text…)Courtesy: Andrew Tomkins

12Yahoo! Research

User generated content (UGC)

• New content– 2 billion pages of editorial content– Tiny number of songs, etc– 5-10 billion pages UGC exist (already ~10%

of consumption), growing

• Note: UGC did not exist as a category a couple of years ago!– Rapidly becoming a key growth area of

consumed web content – but we don’t know how to process it!

Courtesy: Andrew Tomkins

13

Tags The simplest form of UGC

Is the Turing test always the right question?

14Yahoo! Research

15Yahoo! Research

The power of social tagging

• Flickr – community phenomenon

• Millions of users share and tag each others’ photographs (why???)

• The wisdom of the crowd can be used to search

• The principle is not new – anchor text used in “standard” search

• Don’t try to pass the Turing test?

16Yahoo! Research

Anchor text

• When indexing a document D, include anchor text from links pointing to D.

www.ibm.com

Armonk, NY-based computergiant IBM announced today

Joe’s computer hardware linksCompaqHPIBM

Big Blue today announcedrecord profits for the quarter

17Yahoo! Research

Challenges in tag-based search

• How do we use these tags for better search?

• How do you cope with spam?

• What’s the ratings and reputation system?

• The bigger challenge: where else can you exploit the power of the people?

• What are the incentive mechanisms?– Luis von Ahn (CMU): The ESP Game

18Yahoo! Research

19

More UGC: Social search

Indexing the knowledge in people’s heads

20Yahoo! Research

21Yahoo! Research

22Yahoo! Research

Social content Social capital

23Yahoo! Research

Incentives

24Yahoo! Research

Incentives

• What assignment of incentives leads to good user behavior?

– What’s “good” user behavior?

– Good questions, good answers, new questions?

• Whom do you trust and why?

– Propagation of trust and distrust.

25Yahoo! Research

Ratings and reputation

• Node reputation: Given a DAG with– a subset of nodes called GOOD

– another subset called BAD

– Find a measure of goodness for all other nodes.

• Node pair reputation: Given a DAG with a real-valued trust on the edges– Predict a real-valued trust for ordered node

pairs not joined by an edge

Metric labelling

26

CPC advertisements

What pays the bills

27Yahoo! Research

Ads

28Yahoo! Research

Generic questions

• Of the various advertisers for a keyword, which one(s) get shown?

• What do they pay on a click through?

• The answers turn out to draw on insights from microeconomics

29Yahoo! Research

Ads go in slots like this one

and this one.

30Yahoo! Research

Advertisers generally prefer this slot

to this one

to this one

to this one.

31Yahoo! Research

Click through rate r1 = 200 per hour

r2 = 150 per hour

r3 = 100 per hour

etc.

32Yahoo! Research

Why did witbeckappliance win

over ristenbatt?

33Yahoo! Research

First-cut assumption

• Click-through rate depends only on the slot, not on the advertisement

• In fact not true; more on this later.

34Yahoo! Research

Advertiser’s value

• We assume that an advertiser j has a value vj per click through

– Some measure of downstream profit

• Say, click-through followed by• 96% of the time, no purchase

• 0.7% buy Dishwasher, profit $500

• 1.2% buy Vacuum Cleaner, profit $200

• 2.1% buy Cleaning agents, profit $1$ 5.921

35Yahoo! Research

Example

• For the keyword miele, say an advertiser has a value of $10 per click.

• How much should he bid?

• How much should he be charged?

The value of a slot for an advertiser,

what he bids and

what he is charged, may all be different.

36Yahoo! Research

Advertiser’s payoff in ad slot i

(Click-through rate) x (Value per click) – (Payment to search engine)

= ri vj – (Payment to Engine)

= ri vj – pij

Payment ofadvertiser j

in slot iFunction of all other bids.

37Yahoo! Research

Two auction pricing mechanisms

• First price: The winner of the auction is the highest bidder, and pays his bid.

• Second price: The winner is the highest bidder, but pays the second-highest bid.

• Engine decides and announces pricing.

• What should an advertiser bid?

Not truthful.

38Yahoo! Research

Second-price = Vickrey auction

• Consider first a single advt slot

• Winner pays the second-highest bid

• Vickrey: Truth-telling is a dominant strategy for each player (advertiser)

– No incentive to “game” or fake bids

39Yahoo! Research

Auctions and pricing: multiple slots

• Overture’s model:

– Ads displayed in order of decreasing bid

– E.g., if advertiser A bids 10, B bids 2, C bids 4 – order ACB

• How do you price slots? Generalized Vickrey?

– Generalized second-price (GSP)

– Vickrey-Clark-Groves (VCG): each advertiser pays the externality he imposes on others

40Yahoo! Research

Bidder A, $10

Bidder C, $4

Bidder B, $2

Pays 4

Pays 2

Generalized Second Price auction pricing

41Yahoo! Research

VCG pricing

• Suppose click rates are 200 in the top slot, 100 in the second slot

• VCG payment of the second player (C) is 2 x 100 = 200

• For the first player, 4x(200-100) + 200

Externality on third player B.

Externality on C. Externality on B.

42Yahoo! Research

VCG and GSP

• Truth-telling is a dominant strategy under VCG …

• Truth-telling not dominant under GSP!

Edelman, Ostrovsky, Schwarz

43Yahoo! Research

VCG and GSP

• Static equilibrium of GSP is locally envy-free: no advertiser can improve his payoff by exchanging bids with advertiser in slot above.

• Depending on the mechanism, revenue varies: GSP ≥ VCG.

Edelman, Ostrovsky, Schwarz

Locally envy-free mechanisms correspondto Stable Marriage solutions.

44Yahoo! Research

GSP for bid-ordering

• What’s good about bid-ordering and GSP?

–Advertisers like transparency

• What’s wrong with bid-ordering?

45Yahoo! Research

Brand advertising?

Bid ordering(former Yahoo! order)

46Yahoo! Research

47Yahoo! Research

Revenue ordering

• Simplified version of Google’s ordering

– Each ad j has an expected click-through denoted CTRj

– Advertiser j’s bid is denoted bj

• Then, expected revenue from this advertiser is Rj = bj+1 x CTRj

• Order advertisers by Rj

– Payment by GSP

48Yahoo! Research

49Yahoo! Research

“current” Yahoo! ordering

50Yahoo! Research

“Squashed” ordering

• Overture/Old Yahoo! scheme

– Order ads by bid

• Google (puportedly)

– Order by bid click-through rate (CTR)

• Squashing (Lahaie/Pennock)

• Key – advertisers react to mechanism!

s=0 s=1Order by bid*(CTR)sOverture Google?

51Yahoo! Research

52Yahoo! Research

Where do we go next?

• Premise:

– People don’t want to search

– People want to get tasks done

I want to book a vacation in Tuscany.Start Finish

53Yahoo! Research

What is missing?

• Information integration

– Information extraction

– Schema normalization

• Mining social structure

– Tags, UGC

Welcome to The SavoyLocated on The Strand in the heartof the West End theatre district,

hotel near leicester squareSearch

54Yahoo! Research

Computational microeconomics

• Reputation and incentive mechanisms

• Matching marketplaces– Jobs, dates, …

– Online matching everywhere• Hardest part is estimating the payoffs, not the

matching algorithm

• “Network effects”– Are 500 million users 500 times as valuable

as a million users? 5000 times?

55Yahoo! Research

A new convergence

• Monetization and economic value an intrinsic part of system design

– Not an afterthought

– Mistakes are costly!

• Computing meets humanities like never before – sociology, economics, anthropology …

56

Thank you.

Questions?pragh@yahoo-inc.com

http://research.yahoo.com