アイルランドで6月5～7日の日程で行われているICWSM12に参加しているので，
聴講メモを貼り付けておきます．
6/5発表分
Distributional Footprints of Deceptive Product Reviews
Customer's review
5と評価する人もいれば，1を評価する人もいる
1-5の評価分布から，ステマを探す
National vs Distorted(自然？ゆがんでる？)
Find from distribution
SingleReviewかMultiReviewかで行動が違う
Singleだと1か5が多い
Multiは4-5が多い
S-Shape vs J-Shape(Single)
Characterization of rating distribution
D sorted by #
4，2が高いものと4，5が高いものが多い
MultiTimeとSingleTimeで形が全然違う
51・・と15・・が多い
Detection strategies to identify deceptive business entities
SingleReviewer vs MultiReviewer
positive/negative
Temporal boost in ratingの発見
Novel evaluation methodologies
Pseudo-Gold Standard Data
72~75%の判別率
Baselineが40%程度
Privacy in Interaction: Exploring Disclosure and Social Capital in Facebook
Presentation of Self in Everyday Life.
いろいろなクラスタが本人を中心にネットワークを形成しているはず
Social Capital
Access to emotional and substantive support often from strong ties.
Face book user variables
Signals of Relational Investment
Responding to friends' question because,
-Create an expectation about reciprocal behavior
-Perform a social grooming function
-Potentially train the news feed
-Comments on friends update and wall post are seen by the friends' network
Friendの割合が高くなるほど，FriendsOnlyが増えてくる
Bonding SC:Benefits of focused inclusive interaction
Bridging SC:Costs of highly targetting disclosing
Don’t Disturb My Circles! Boundary Preservation Is at the Center of Location-Sharing Concerns
What is the motivation of Boundary regulation
Online Boundary Regulation
preserve offline relation ship
Friendに対してはPrivacyを考える
Familyは考えない
Strangerに対しても考えない(日本とは違うなあ)
Boundary preservation concern = BPC
Facebook and Privacy: The Balancing Act of Personality, Gender, and Relationship Currency
sentiment的に友人と自分には相関がある
disclose & conceal what is considered private and public in Facebook
Count number of disclosed & concealed from facebook properties
Item Response Theory(IRT)
P_IJ= 1/(1+exp(-a_i(t_j-B_i)))
1. Smart Privacy Mob
About privacy, Westine has divided people in
1.privacy fundamentalists
2.pragmatic majority
3.marginally concerned
2. Who are they?
Those who share more sensitive info are
-open to new experience ------important
-self monitoring
-male --------------------------important
-more active
-younger
3. What's sensitive
position, employer
residence, hometown not sensitive in circle.
The Livehoods Project: Utilizing Social Media to Understand the Dynamics of a City
neighborhoods
Two perspectives
Politically constructed 政治的に作られた構造か
Socially constructed 社会的に作られた構造か
Collective Cognitive Map
どこでTweetしたかでユーザをグループ分け
そこから，土地のグループ分け．
Tweetにも応用可能か．
場所の類似性(Jaccard係数)
How to evaluate
実際にピッツバーグの人に聞いてみた
Who Does What on the Web: A Large-Scale Study of Browsing Behavior
Nielsen MegaPanel
265000 individuals in US
user data, log of anonymized complete browsing activity from 2009/6-2010/5
Younger and more educated => more use web, more active
色々調べているけど・・・
education が大きな効果を持つ
Evolution of Experts in Question Answering Communities
Expert Influence best answer selection
Ordinary users get intimated by expert
Experts avoid other experts
Experts evolve with differente patterns
These experes can be found with satisfactory performance within 20 weeks
Temporal Motifs Reveal the Dynamics of Editor Interactions in Wikipedia
Motifを使ってWikipediaの編集を分析
編集合戦中とかMotifから判別可能
Modeling Spread of Disease from Social Interactions Best Paper Candidate!
SVMでインフルエンザであるTweetかどうかを判断
Modeling Destructive Group Dynamics in On-line Gaming Communities
WoWを対象にグループの生成と解散を分析
ギルドの歴史から分析可能
Charactor&Guildの二部ネットワーク
Guild抜けは感染する
Quitting Eventを予測可能か？
Quitting Eventが起きないことは90％近い予測
Quitting Eventが起きる可能性は30%程度
ソーシャルネットワークとして調べると面白い
GroupDynamicsが直接見えるデータというのは貴重だ．
投稿
