• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
[International Asian LOD Challenge Day 2012]LOD generation of Social and Mass media data: Apply to media comparisons
 

[International Asian LOD Challenge Day 2012]LOD generation of Social and Mass media data: Apply to media comparisons

on

  • 423 views

I'll present at Nara 1st Dec, 2012

I'll present at Nara 1st Dec, 2012

Statistics

Views

Total Views
423
Views on SlideShare
423
Embed Views
0

Actions

Likes
0
Downloads
4
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    [International Asian LOD Challenge Day 2012]LOD generation of Social and Mass media data: Apply to media comparisons [International Asian LOD Challenge Day 2012]LOD generation of Social and Mass media data: Apply to media comparisons Presentation Transcript

    • You  can  download  this  slide  here:      hHp://slidesha.re/TwzFrf     LOD  genera*on   of  Social  and  Mass  media  data:   Apply  to  media  comparisons Interna*onal  Asian  LOD  Challenge  Day   1st  Dec,  2012   Presenter:  Kenji  Koshikawa   Co-­‐Researcher(Adviser):  T.  Kawamura,    H.  Nakagawa,  Y.  Tanaka,  A.  Ohsuga   Affilia*on:  Department  of  Social  Intelligence  and  InformaBcs   Graduate  course  of  InformaBon  Systems   The  University  of  Electro-­‐CommunicaBons    
    • About  Our  Project 2
    • Project  Abstract  We  do  just  two  things  on  the  project:  1.   Building  seman*c  networks      from  media  informa*on  2.   Comparing  with  different  media      using  the  networks.   3
    • Represen*ng  events  informa*on   using  seman*c  network  (RDF)  1/2 Example1: 昨日太郎は秋葉原でiPhone5を購入したので、幸せそうだった。  (Yesterday,  Taro  bought  a  iPhone  5  at  Akihabara,  so  he  looked  happy.) Event  1 Event  2 Conver*ng  natural  language  into  seman*c  networks Cause Event  2 Event  1 太郎(Taro) 秋葉原   Ac*vity Status (Akihabara)  Loca*on   購買  Time 昨日   Time 幸福   Object (Buying) (Yesterday) (Happiness)   iPhone  5 4
    • Outpu[ng  Linked  Data  as  RDF/XML  format   e.g.  “Taro  bought  a  iPhone  5  at  Akihabara,  so  he  looked  happy.” 5
    • Represen*ng  events  informa*on   using  seman*c  network  (RDF)  2/2 Example  2  (from  real  media):  a  fall  accident April  an  accident  to  occur  the  southern  state  a  poor  maintenance of  Florida  June  the  state  of  Florida,  U.S. 6
    • Project  Abstract  We  do  just  two  things  on  the  project:  1.   Building  seman*c  networks      from  media  informa*on  2.   Comparing  with  different  media      using  the  networks.   Mass  media Social  media 7
    • A Case of media comparison Topic: Introduction of Osprey in Japan About Dataset : Period: 1st April – 16th Aug, 2012 Condition: Media textual information have a word “オスプレイ”(Osprey). Dataset of Social media: Twitter: 3,084 tweets A  photo  of  Osprey Dataset of Mass media: Asahi digital news paper: 116 articles MSN Sankei news: 231 articles Nippon News Network(NNN): 110 articles Fuji News Network(FNN): 78 articles
    • Consideration throughout visualizingnetwork •  the  difference  of  diversity  of  topic      between  each  media     •   easy  to  access  minority  opinion   •   the  existence  of  2  kinds  of  osprey  (introduce)   •   the  Laterality  of  dependence  on      user  loca*on   9
    • Summary  of   the  existence  of  2  kinds  of  osprey On mass media there are NOT information about following: •  The existence of other variants (of Osprey) •  The relation between the variants and the accident rate •  The fact that the accident rate of a variant, be deployed in Japan is Lower than other rotorcraft ※ ※ The  V-­‐22s  accident  rate  is  the  lowest  of  any  Marine  rotorcrab  [Ref  01] By visualizing, we found the existence of 2 kinds of osprey andthe relation between the variants and accident rate. Thus, we could notice a doubt of media bias on mass media. 
A doubt of media bias “Mass media hardly report about such information intentionally, and they was in a mood in the press fomenting the contrary opinion about introduction of osprey in Japan.” 10
    • Example  of  Considera*on:    the  existence  of  2  kinds  of  osprey Look  around  a  “deploying”  node deploying CV-­‐22  osprey A Color of node means the occurrence rate on each media. Social Mass MV-­‐22  osprey a common concept This Figure has been showing that there are 2 kinds of variants of osprey according to the network built by social media dataset. 11
    • Example  of  Considera*on:    the  existence  of  2  kinds  of  osprey CV-­‐22  Osprey deploying Be  nothing  like MV-­‐22  Osprey Lower for  transport,  original  requirement Harmful  rumor There are the difference of use of each variant of osprey, It can be read from this figure. e.g. MV-22: for transporting / CV-22: for ? 12
    • Example  of  Considera*on:    the  existence  of  2  kinds  of  osprey Accident  rate Copter low Pilot  error Look  around  a  “accident  rate”  node 13
    • Look around Example  of     a “Accident rate of Osprey” node Considera*on:   the  existence  of     Low 2  kinds  of  osprey Accident  rate  of  Osprey Look around a “1.93” node Look around a “13.47” node Accident  rate Accident  rate Accident  rate  of  CV-­‐22 Accident  rate  of  MV-­‐22 for  the  Special  Opera*ons  Command Accident  rate  of  Osprey Accident  rate  of  Osprey 14
    • Look around Example  of     a “Accident rate of Osprey” node Considera*on:   the  existence  of     Low 2  kinds  of  osprey The  rela*on  between  the  variants  ate  of  Osprey Accident  r and  Look around a “1.93” node reflected.    (from  a ocial   node  the  accident  rate  was   Look around s “13.47” Accident  rate media  dataset) Accident  rate Accident  rate  of  CV-­‐22 Accident  rate  of  MV-­‐22 for  the  Special  Opera*ons  Command Accident  rate  of  Osprey Accident  rate  of  Osprey 15
    • Summary •  Introduced  our  project:   –  To  generate  LOD  from  media  informa*on   –  To  compare  with  different  media  using  the  Linked  Data  •  We  are  looking  for  solving  below:   –  en*ty  resolu*on,  instance  matching  problem   –  connect  to  other  Linked  Data  •  In  future  work,  we  will  concentrate  on  improving    LOD  visualiza*on   for  knowledge  discovery.  •  If  you  know  interes*ng  topic  for  media  comparison,  let  me  know.     16
    • Reference [Ref  01]    "V-­‐22  Is  The  Safest,  Most  Survivable  Rotorcrab  The   Marines  Have."LexingtonInsBtute.org,  February   2011.  Retrieved:  16  February  2011.    [Ref  02]  (Japanese)   越川 兼地,  川村 隆浩,  中川 博之,  田原 康之,  大須賀 昭彦:  CRFを用いた メディア情報の抽出とLinkedData化 -­‐  ソーシャルメディアとマスメディアの 比較事例 -­‐  ,合同エージェントワークショップ&シンポジウム(JAWS  2012),   2012.   Slide  (wriHen  in  Japanese):    hHp://slidesha.re/11pf0qR  
    • Appendix
    • Goal  /  Mo*va*on  1.  To  generate  Linked  Data  from  Media   Informa*on   –  Mo*va*on:   •  to  organize  abundance  informa*on     •    to  make  us  recognize  real  events  easily2.  To  compare  with  different  media  using  the   Linked  Data  (we  generated)   –  Mo*va*on:     •  to  discover  knowledge  from  the  difference  of  informa*on  between   media   •  to  understand  real  events  from  mul*ple  points  of  view 19
    • Our  System  Overview 20
    • Visualizing  the  NetworkSize of node/Thickness of edge: are calculated based on the frequency information. Color of node: expresses the occurrence rate of Social Mass concept between each media a common using 5 colors. concept Color of edge: expresses kind of relationship between two concepts. subject object time status quoted source activity location target cause ※we used a visualization Application: Gephi 0.8.1 beta 21
    • Future  Work   •  At  this  stage  we  just  visualize  the  network,  so  users  have  to   discover  knowledge  themselves.   –  We  are  developing  tools  to  support  for  knowledge  discovery  from  the   network.   •  To  es*mate  important  node/sub-­‐network  in  the  network.    •  to  evaluate  our  system  and    to  be  needed  to  experience  other  topic  •  We  are  looking  for  solving  below:   –  en*ty  resolu*on,  Instance  matching.    •  We  will  go  up  for  LOD  Challenge  2012  Japan.   –  But,  I’m  not  sure  which  sec*on  is  the  best  for  our  project. Dataset Idea Applica*on Visualiza*on 22
    • 整理:  MV-22  /  CV-22 英語にする オスプレイの型番と事故率の関係 型番 用途 事故率 MV-­‐22   輸送用 1.93    (日本配備) 米海兵隊所属   -­‐ 2.45 航空機平均 CV-­‐22 特殊作戦用(空軍) 13.47 日本に配備される(た)機種 「MV-22」の事故率は低い. 23
    • 事象の表現方法   事象情報を表現するために,[Nguyen 12]の   行動属性を拡張し9つの事象属性を定義した. Event  descripDon   describe property Subject Subject  of  an  event Ac*vity Ac*vity  of  an  event   Object Object  of  an  ac*vity Target  (new) Against  whom  (e.g.  people,  country,  …) Status(new) Status  of  a  subject Loca*on Loca*on  where  an  event  occurred Time Time  informa*on  when  an  event  occured Cause  (new) Cause  what  an  event  occurred Quoted  source  (new) Source  of  a  quote [Nguyen  12]   The-­‐Minh  Nguyen,  Takahiro  Kawamura,  Yasuyuki  Tahara,    and    Akihiko  Ohsuga:  Self-­‐Supervised  Capturing  of  Users’  Ac*vi*es  from   24 Weblogs.  Interna*onal  Journal  of  Intelligent  Informa*on  and  Database  Systems,Vol.6,  No.1,  pp.61-­‐76,  InderScience  Publishers,  2012
    • End