TwiNews: Detect local user interests in Twitter to re- rank news Xin Shuai Tuesday, October 8, 2013
Insight demo
Oct 28, 2014
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
TwiNews: Detect local user interests in Twitter to re-
rank newsXin Shuai
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Top ranked news
may not satisfy all users
from different geographicallocations!
Tuesday, October 8, 2013
Not relevant to
China!
Tuesday, October 8, 2013
Word Closeness Local User Interests
Tuesday, October 8, 2013
How to detect the local user interests given a news query?
Tuesday, October 8, 2013
How to detect the local user interests given a news query?
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Can we provide localized news ranking for users by leveraging
geographical tweets?
Tuesday, October 8, 2013
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
tweet 1
tweet 2
tweet n
topic 1 topic 2
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
tweet 1
tweet 2
tweet n
topic 1 topic 2
local topical distribution
+
+
+technology (topic 1)
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
tweet 1
tweet 2
tweet n
topic 1 topic 2
local topical distribution
+
+
+technology (topic 1)
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
tweet 1
tweet 2
tweet n
topic 1 topic 2
news 1
news 2
news 3
technology (topic 1)
local topical distribution
+
+
+technology (topic 1)
Tuesday, October 8, 2013
LDAtopic model
topic 1 technology:
mobileiphone
topic 2economy:
marketstock
investor
tweet 1
tweet 2
tweet n
topic 1 topic 2
news 1
news 2
news 3
technology (topic 1)
local topical distribution
+
+
+technology (topic 1)
cosine similarity
cosine similarity
cosine similarity
Tuesday, October 8, 2013
A/B Testing
TwiNews
30 queries, 2 weeks100 turkers from CA & NY
+3%
+7%
t-test: p < 0.05
A
B
Tuesday, October 8, 2013
GibbsLDA++
Tuesday, October 8, 2013
localized news ranking
localized user interests from Twitter
Tuesday, October 8, 2013
localized news ranking localized targeted advertising
localized user interests from Twitter
Tuesday, October 8, 2013
localized news ranking localized targeted advertising
localized deal recommendation
localized user interests from Twitter
Tuesday, October 8, 2013
localized news ranking localized targeted advertising
localized deal recommendation
localized user interests from Twitter
localized music/movie recommendation
Tuesday, October 8, 2013
Xin Shuai
Tuesday, October 8, 2013
LDA topic modelingLDA (Latent Dirichlet Allocation)
Tuesday, October 8, 2013
LDA topic modelingLDA (Latent Dirichlet Allocation)
topic 1:mobileiphone
facebookipad
topic 2:marketstockindex
investor
topic 3:............
Tuesday, October 8, 2013
Amazon Mechanical TurkPrice: eleven cents/HITS
Total Cost: over 200 dollars
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Tuesday, October 8, 2013
Related Documents
