�ݺ�ߣ

KDD
KDD’
’10 Tutorial:
10 Tutorial:
Recommender Problems for
Recommender Problems for
Web Applications
Web Applications
Deepak Agarwal and Bee-Chung Chen
Yahoo! Research

2
Deepak Agarwal & Bee-Chung Chen @ KDD’10
Agenda
• Focus:
– Recommender problems for dynamic
dynamic, time
time-
-sensitive
sensitive applications
• Content Optimization
• Introduction (20 min, Deepak)
– Content optimization, match-making, example applications
• Offline components (40 min, Deepak)
– Collaborative filtering (CF), methods for cold-start
• Online components + initialization (70 min, Bee-Chung)
– Time-series, online/incremental methods, explore/exploit (bandit)
• Evaluation methods (15 min, Deepak)
• Challenges (5 min, Deepak)

3
Content Optimization
• Goal
– Effectively and “pro-actively” learn from user interactions with
content that are displayed to maximize our objectives.
• A new scientific discipline at the interface of
– Large scale Machine Learning & Statistics
• Offline Models
• Online Models
• Collaborative Filtering
• Explore/Exploit
– Multi-Objective Optimization in the presence of Uncertainty
• Click-rates (CTR), Engagement,….
– User Understanding
• Profile construction
– Content Understanding
• Topics, “aboutness”, entities, follow-up of something, breaking news,…

4
Content Optimization: High level flowchart
• Flow
– Understand content (Offline)
– Serve content to optimize our objectives (Online)
– quickly learn from feedback obtained using ML/Statistics (Offline +
Online)
– Constantly enhance our content inventory to improve future
performance (Offline)
– Constantly enhance our user understanding to improve future
performance (Offline + Online)
– Iterate

5
Some examples
• Simple version
– I have an important module on my page, content inventory is
obtained from a third party source which is further refined through
editorial oversight. Can I algorithmically recommend content on this
module? I want to drive up total CTR on this module
• More advanced
– I got X% lift in CTR. But I have additional information on other
downstream utilities (e.g. dwell time). Can I increase downstream
utility without losing too many clicks?
• Highly advanced
– There are multiple modules running on my website. How do I take
a holistic approach and perform a simultaneous optimization?

6
Recommend applications
Recommend search queries
Recommend news article
Recommend packages:
Image
Title, summary
Links to other pages
Pick 4 out of a pool of K
K = 20 ~ 40
Dynamic
Routes traffic other pages

7
Problems in this example
• Optimize CTR on different modules together in a holistic
way
– Today Module, Trending Now, Personal Assistant, News, Ads
– Treat them as independent?
• For a given module
– Optimize some combination of CTR, downstream engagement and
perhaps revenue.

8
Single module CTR optimization problem
• Display “best” articles for each user visit
• Best - Maximize User Satisfaction, Engagement
– BUT Hard to obtain quick feedback to measure these
• Approximation
– Maximize utility based on immediate feedback (click rate) subject
to constraints (relevance, freshness, diversity)
• Inventory of articles?
– Created by human editors
– Small pool (30-50 articles) but refreshes periodically

9
Recommendation: A Match-making Problem
Opportunity
Users, queries,
pages, …
Item Inventory
Articles, web page,
ads, …
Use an automated algorithm
to select item(s) to show
Get feedback (click, time spent,..)
Refine the models
Repeat (large number of times)
Measure metric(s) of interest
(Total clicks, Total revenue,…)
• Recommendation problems
• Search: Web, Vertical
• Online advertising
• …

10
• Items: Articles, web pages, ads, modules, queries, users, updates, etc.
• Opportunities: Users, query keywords, pages, etc.
• Metric (e.g., editorial score, CTR, revenue, engagement)
– Currently, most applications are single-objective
– May be multi-objective optimization (maximize X subject to Y, Z,..)
• Properties of the item pool
– Size (e.g., all web pages vs. 40 stories)
– Quality of the pool (e.g., anything vs. editorially selected)
– Lifetime (e.g., mostly old items vs. mostly new items)
Important Factors affecting solution in Match-making Problems

11
Factors affecting Solution continued
• Properties of the opportunities
– Pull: Specified by explicit, user-driven query (e.g., keywords, a form)
– Push: Specified by implicit context (e.g., a page, a user, a session)
– Size (e.g., user base); continuity (e.g., session vs. single event)
• Properties of the feedback on the matches made
– Types and semantics of feedback (e.g., click, vote)
– Latency (e.g., available in 5 minutes vs. 1 day)
– Volume (e.g., 100K per day vs. 300M per day)
• Constraints specifying legitimate matches (e.g., business
rules)
• Available Metadata (e.g., link graph, various user/item attributes)

12
Recommendation vs. Other Match-Making Problems
Ads
Anything
(except for ads)
Anything
(except for ads)
Items
Sponsored search
Content match
Behavior targeting
Display advertising
(non-guaranteed)
Web search
Vertical search
Recommend articles,
friends, feeds to
users
Recommend related
items given an item
Examples
Push
Pull (explicit)
Users specify their info
needs
Push (implicit)
The system guesses
users info needs
Opportunities
Revenue
Relevance to the query
User engagement
Main Metric
Advertising
Search
Recommendation

13
More on Recommendation vs. Search
• Recommendation
– User intent: See something “interesting” (browse mode, implicit)
• The system tries to guess what a user likes on an entity/topic
page
– No query reformulation (unless we suggest related topics/entities)
– False +ve more costly than false –ve
• Showing a bad article is a worse than missing a good one
• Search
– User intent: Explicit, users express what they want
– Users can reformulate queries
– False –ve more costly (but depends on the query)
• Users want to get the results they are looking for

14
Modeling: Key components
Offline
(Logistic, GBDT,..)
Feature construction
Content: IR, clustering, taxonomy, entity,..
User profiles: clicks, views, social, community,..
Online
(Fine resolution
Corrections)
(item, user level)
(Quick updates)
Explore/Exploit
(Adaptive sampling)
Initialize

15
Modeling Problems that has received attention
• Univariate response (e.g. click); single objective (e.g.
maximize CTR)
• Our solution
– Initialize online through offline models
– Learn “corrections” to offline models at very granular levels (user,
item) and learn rapidly in an online fashion
– Online correction models have reduced dimension through clever
representations of parameters and by exploiting the fallback
mechanism to coarser models
– The models are tightly coupled with Explore-exploit to ensure fast
convergence to areas of high valued response

Example Application:
Example Application:
Today Module on Yahoo! Homepage
Today Module on Yahoo! Homepage
Currently in production powered by some methods
discussed in this tutorial

17
Recommend packages:
Image
Title, summary
Links to other pages
Pick 4 out of a pool of K
K = 20 ~ 40
Dynamic
Routes traffic other pages
1
1 2
2 3
3 4
4

18
Problem definition
• Display “best” articles for each user visit
• Best - Maximize User Satisfaction, Engagement
– BUT Hard to obtain quick feedback to measure these
• Approximation
– Maximize utility based on immediate feedback (click rate) subject
to constraints (relevance, freshness, diversity)
• Inventory of articles?
– Created by human editors
– Small pool (30-50 articles) but refreshes periodically

19
Where are we today?
• Before this research
– Articles created and selected for display by editors
• After this research
– Article placement done through statistical models
• How successful ?
"Just look at our homepage, for example. Since we began pairing our content
optimization technology with editorial expertise, we've seen click-through rates
in the Today module more than double. ----- Carol Bartz, CEO Yahoo! Inc (Q4,
2009)

20
Main Goals
• Methods to select most popular articles
– This was done by editors before
• Provide personalized article selection
– Based on user covariates
– Based on per user behavior
• Scalability: Methods to generalize in small traffic scenarios
– Today module part of most Y! portals around the world
– Also syndicated to sources like Y! Mail, Y! IM etc

21
Similar applications
• Goal: Use same methods for selecting most popular, personalization
across different applications at Y!
• Good news! Methods generalize, already in use

22
Next few hours
Prior estimation,
dimension reduction
Prior estimation
Intelligent Initialization
Bandits with covariates
Multi-armed bandits
Explore/Exploit
Incremental CF,
online regression
Time-series models
Online Models
Collaborative filtering
(cold-start problem)
Offline Models
Personalized
Recommendation
Most Popular
Recommendation

�ݺ�ߣ

Recommender Problems Introduction

More Related Content

Recommender Problems Introduction