Date post: | 24-May-2015 |
Category: |
Technology |
Upload: | terrierteam |
View: | 261 times |
Download: | 1 times |
The Horizons of News SearchSSM Workshop 2010By Richard McCreadie, Craig Macdonald, Iadh Ounis
1
How is News reported?
2
Observer
The public witnesses event
Reporter
Reporters arrive and write an article
Internet
NewswireCompany
Newswire companies publish
the article
Observers now report events online
Both publicly
Instant Messaging Logs
And implicitly . . .
News is commentated on and
discussed . . .
. . .. . .
. . .
Event!
How do we access that news?
• Integrated News Search– – – – etc . . .
• News Aggregators– – – – etc . . .
3
Motivations for Improving News Search
• Improving real-time search– Detect events more quickly (freshness)– Classify user queries more accurately– Attain a better coverage of news stories– Enable the serving of a wider definition
of `news’
• Improve coverage in news results– Avoid institutional bias of newswire
companies– Provide a sample of `public opinion’
4
What’s the latest news from the SSM
Workshop?
*** ***
How can User-Generated Content Help?
5
Detect news-related queries for recent news stories for which there aren’t any news articles
Improve confidence in
classifying queries based
on news discussion in
user –generated content
Display `first source’ for a news story
Provides content for new news
articles with few prior articles
Provide a sample of
`public opinion’ about a news
story
Mitigates institutional bias in news articles
Freshness Volume
ClassifyingUser Queries
RankingResults
Challenges to be addressed
• Confidence– How can we know when a source is
trustworthy?– Inaccuracies/falsification/spam?
• Quality– Does user-generated content really add
value?– How to distinguish `good’ posts?– What sources should we include?
• Coverage– What stories class as news?– How to distinguish between real events
and `chat’?– Should results be diversified?
7
?
?
?
?
vs
Outlook
• Users now act as both reporters, commentators and consumers of news
• This creates exciting possibilities in leveraging the rich information they produce to improve news search
• CR: TREC Blog track, Top News Stories Identification
8
News