+
360° Social Media Analysis (BETA)
|
+
360° Social Media Analysis (BETA)
|
|
Can #socialmedia give deeper insights into #election2012 based on #sentiment #emotion and #popularity analysis/cues?
How do events (e.g., debates, Bengazi attack) unfold through sentiments, emotions and popularity associated with candidates, and what thematic insights (reasons) do they reveal?
Below are just some of the examples of unusual insights associated with Sentiment, Emotion and Popularity that you will find at Twitris live: http://twitris.knoesis.org/election/sentiment/ . [Note: use date selection field to go back in time.]
Send us your questions/feedback at amit@knoesis.org
Sentiment Analysis on Election Day
The following screenshots were taken on Nov. 6th 6pm. Twitris shows that Obama leads the race in Colorado, Florida, Iowa and Ohio, while Obama/Romney race is very close in Virginia.
Sentiment Analysis on Presidential Debate
Obama
Romney
Twitris shows the negative topics related to the decrease on Oct. 14th including:
supporter wears shocking racist t-shirt, romney becomes president,
sensata workers are living proof, hits back hard at romney, etc.
And positive topics related to the increase on Oct. 15 such as whose website is
faster, mitt romney 's website, more than two seconds.
Emotion
Emotion Joy:
Twitter users got excited before both debates on Oct. 3rd and Oct. 16th. For example, “1st debate tonight, Obama v #Romneyshambles should be fun”, “Looking forward to the 2nd presidential debate tonight . . . unless Obama pulls another Norv.”
Twitter users were also very excited when Obama went to daily show. For example, “PRESIDENT OBAMA WILL BE ON THE DAILY SHOW TONIGHT AT 11PM EVERYBODY WATCH!”
Emotion Anger:
Twitter users were angry about Romney on Sep. 24, 25 and
26, because of the “airplane
windows” incident. I.e., “Mitt
Romney literally said he doesn't understand why airplane windows
don't open? And we're still letting him run for
president?”
Twitter users were angry about Romney on Oct. 5, 6 and 7, because of the “big bird” incident. I.e., “Mitt Romney's Pathetic. How Can He Cancel Big bird ? Like wtf. Sit Your Ass White Ass Down.”
Emotion Fear:
There is a pike on emotion fear on Sep. 26th. And the reason is that the US embassy in Libya was attacked on that day.
Popularity
| Voter's Rank | Candidate | Trending Causal Tweets | |||||
|---|---|---|---|---|---|---|---|
| Nov. 6, 2012, 3:30pm EST | |||||||
| 1 | Barack Obama |
|
|||||
| 2 | Mitt Romney |
|
|||||
Analysis: Twitter Interactions in topical communities during US Election 2012
Period: Oct 1 to Nov. 3, 2012
Corpus: 4 M tweets and 1.9 M users in interactions
Observations:
Oct. 3, 2012.
The following interaction networks of influential users talking about Romney vs. Obama show the user's sentiment for the target candidate (e.g., Positive attribute in the Obama's network shows positive leaning of the influencer to Obama side).
1.) Before the first presidential debate, the structure of interaction network of top 100 influencers in the two communities looked like following-
Romney's community before 1st debate
|
|
Obama's community before 1st debate
|
|
|
Oct. 16, 2012.
2.) After the first presidential debate, the structure of interaction network of top 100 influencers in the two topical communities took completely different shape:
Romney's community after 1st debate
|
|
Obama's community after 1st debate
|
|
|
Oct. 18, 2012.
3.) After the second presidential debate, the structure of interaction network of top 100 influencers in the two topical communities again showed surprising shapes even though President Obama performed really well, does it mean that Gallup is getting it right?
Romney's community after 2nd debate
|
|
Obama's community after 2nd debate
|
|
|
Oct. 21, 2012.
4.) After the second presidential debate, the structure of interaction network of top 100 influencers in the two topical communities showed evolution in the cohesiveness for President Obama, mostly due to his outstanding performance?
Romney's community after 2nd debate, 3 days
|
|
Obama's community after 2nd debate, 3 days
|
|
|
Oct. 23, 2012.
5.) After the final presidential debate, the structure of interaction network of top 100 influencers in the two topical communities provides insights about increased positive cohesiveness for President Obama:
Romney's community during and after 3rd debate, 14 hours
|
|
Obama's community during and after 3rd debate, 14 hours
|
|
|
Nov. 4, 2012.
5.) After the Hurricane Sandy Storm, the structure of interaction network of top 100 influencers in the two topical communities provides insights about increased positive cohesiveness for President Obama:
Romney's community after Hurricane Sandy
|
|
Obama's community after Hurricane Sandy
|
|
|
| Rank | Candidate | @Media | |
|---|---|---|---|
| May 9, 2012 | |||
| 1 | Barack Obama | 272118 | 39973 |
| 2 | Mitt Romney | 90181 | 19075 |
Twitris uses two indicators extracted from the tweet corpus to forecast the election: (1) popularity of the candidates; (2) positive sentiment of the candidates. Directly using these two indicators have been shown to achieve 80%+ accuracy in the forecast. A more indepth study has been conducted to examine the predictive power of different user groups in predicting the results of Super Tuesday races in 10 states.
The forecast is based on the analysis of tweets in four weeks before the primary day. If the indicators extracted from the four-week data are too close to tell the candidates apart, the indicators extracted from the tweets in the final week before the primary day are employed (e.g, Alabama, Idaho, Ohio).
Bold font highlights the winner. Green background color indicates that the results are consistent with Twitris analysis.
The analysis could not be performed for some candidates in some states (labeled with star*), due to the lack of tweets.
 |
 |
 |
 |
|
| Newt Gingrich | Ron Paul | Mitt Romney | Rick Santorum |
| popularity | positive sentiment |
popularity | positive sentiment |
popularity | positive sentiment |
popularity | positive sentiment |
||
| June 5 | California | 3.9% | 55.2% | 9.1% | 61.4% | 42.6% | 47.5% | 5.1% | 47.9% |
| Montana* | |||||||||
| New Jersey | 4.2% | 54.4% | 7.0% | 62.0% | 44.9% | 48.8% | 6.1% | 53.0% | |
| New Mexico | 7.5% | 49.9% | 12.7% | 50.5% | 36.0% | 44.7% | 7.5% | 49.9% | |
| South Dakota* | |||||||||
| May 29 | Texas | 4.8% | 54.8% | 12.1% | 62.7% | 39.8% | 48.3% | 7.5% | 60.2% |
| May 22 | Arkansas | 6.2% | 33.9% | 13.3% | 54.5% | 45.0% | 42.1% | 28.7% | 41.4% |
| Kentucky | 10.3% | 59.7% | 10.1% | 58.6% | 44.1% | 47.6% | 18.1% | 40.0% | |
| May 15 | Nebraska* | ||||||||
| Oregon | 8.6% | 48.3% | 8.5% | 47.7% | 52.8% | 47.4% | 25.3% | 47.1% | |
| May 8 | Indiana* | ||||||||
| North Carolina | 11.8% | 61.5% | 11.0% | 64.7% | 51.9% | 52.4% | 26.1% | 57.6% | |
| West Virginia* | |||||||||
| April 24 | Connecticut | 9.8% | 51.3% | 14.8% | 67.2% | 55.3% | 55.1% | 31.6% | 56.1% |
| Delaware | 15.3% | 58.8% | 14.5% | 62.9% | 48.5% | 50.8% | 29.7% | 46.0% | |
| New York | 10.2% | 60.5% | 13.6% | 58.3% | 52.8% | 53.9% | 31.2% | 49.0% | |
| Pennsylvania | 11.8% | 47.0% | * | * | 46.0% | 68.6% | 50.5% | 57.9% | |
| Rhode Island | * | * | * | * | 49.2% | 57.2% | 35.9% | 58.3% | |
| April 3 | Wisconsin | * | * | * | * | 77.4% | 55.5% | * | * |
| Maryland | 14.0% | 55.7% | 10.6% | 63.4% | 51.6% | 61.8% | 36.2% | 55.7% | |
| District of Columbia | 12.0% | 55.3% | 12.0% | 57.9% | 39.5% 54.2% |
55.1% 52.1% |
36.7% 31.3% |
51.4% 44.4% |
|
| March 24 | Louisiana | 15.5% | 58.0% | 12.9% | 56.1% | 44.2% | 54.1% | 37.4% | 54.9% |
| March 20 | Illinois | 11.7% | 58.2% | 13.1% | 54.2% | 38.8% | 54.5% | 36.4% | 50.6% |
| March 18 | Puerto Rico* | ||||||||
| March 13 | Alabama | 20.0% | 62.5% | 12.0% | 58.1% | 34.8% 41.6% |
53.7% 66.5% |
34.9% 31.0% |
51.8% 65.7% |
| Hawaii | 25.0% | 62.1% | 19.7% | 63.3% | 31.2% | 61.0% | 24.7% | 47.7% | |
| Mississippi* | |||||||||
| March 10 | Kansas | 15.8% | 61.8% | 13.6% | 55.7% | 38.2% | 56.3% | 31.9% | 46.3% |
| March 6 | Alaska* | ||||||||
| Georgia | 15.6% | 57.5% | 13.1% | 57.4% | 35.3% | 57.7% | 34.7% | 49.1% | |
| Idaho | 10.0% | 58.2% | 15.6% | 70.3% | 38.6% 49.8% |
56.2% 67.3% |
36.0% 19.5% |
59.4% 65.9% |
|
| Massachusetts* | |||||||||
| North Dakota* | |||||||||
| Ohio | 9.4% | 55.0% | 13.8% | 59.7% | 35.5% 41.8% |
55.6% 62.2% |
38.1% 42.4% |
51.7% 42.3% |
|
| Oklahoma | 13.3% | 57.8% | 15.8% | 58.7% | 27.6% | 52.7% | 37.7% | 52.6% | |
| Tennessee | 17.0% | 70.4% | 10.4% | 61.5% | 32.9% | 43.6% | 36.4% | 58.8% | |
| Vermont* | |||||||||
| Virginia* | |||||||||
| Wyoming* | |||||||||
| March 3 | Washington | 11.6% | 57.7% | 12.7% | 58.3% | 29.4% | 50.3% | 36.4% | 48.8% |
| Feb. 28 | Arizona | 12.3% | 54.9% | 17.5% | 59.9% | 37.6% | 52.7% | 26.9% | 51.4% |
| Michigan | 10.2% | 54.5% | 14.9% | 53.5% | 38.3% | 50.1% | 29.7% | 47.0% |
Analysis: Twitter Interactions in topical communities during US Election 2012 event
Period: March 1 to March 31, 2012
Corpus: 4.2 M tweets and 1.1 M users in interactions
Observations:
1.) Both Romney and Santorum topical communities showed presence of a core connected network among influential users over the period of Mar 1 to Mar 31, 2012, which in turn depicts the potential to drive the actions of the community users. Interesting to note is that both communities have core set size nearly equal, 40 (non overlapping users) out of top 100 influencers.
2.) Romney community seems to get better organized and connected in small groups over the time as we move towards the end of March, which is likely due to his winning chances after the Super Tuesday and Illinois primary victory. It can be seen from the jump in community density and modularity for the Romney community. Figure 1 shows the changes in the community clusters and table 1 shows various statistics for the People-Content-Network analysis.
3.) Another interesting point to note here is that though both communities showed denser network in the snapshot-3, still the modularity kept decreasing for Santorum community as compared to Romney, which again points to shifting towards better organization in the Romney community influencers.
4.) We analyzed the common set of influential users in the two communities and found only near 10% shift in their political standing, which is likely due to political affinity of users, leading to bias for a particular candidate and party.
We are investigating further role of events occured during the snapshots, which probably caused the Santorum community less organized as compared to Romney community, so please stay tuned for more exciting insights in our next release!
Table 1: Results of PCNA for topical communities surrounding
Mitt Romney and Rick Santorum
Group formation in the Communities as function of snapshots:
A.) Romney Influencer Community: shifted towards better modularity and strong connectedness over the time.
.. 
.. 
B.) Santorum Influencer Community: shifted initially for better clustering but ultimately poor modularity and connectedness over the time in comparison to Romney community.
.. 
.. 
|
Click on a graph point to see the popular topics for that day.
|
A Semantic Social Web application with real-time monitoring and multi-faceted analysis of social signals to provide insights and a framework for situational awareness, in-depth event analysis and coordination, emergency response aid, reputation management etc.
Users are sharing voluminous social data (800M+ active Facebook users, 1B+ tweets/week) through social networking platforms accessible by Web and increasingly via mobile devices. This gives unprecedented opportunity to decision makers-- from corporate analysts to coordinators during emergencies, to answer questions or take actions related to a broad variety of activities and situations: who should they really engage with, how to prioritize posts for actions in the voluminous data stream, what are the needs and who are the resource providers in emergency event, how is corporate brand performing, and does the customer support adequately serve the needs while managing corporate reputation etc. We demonstrate these capabilities using Twitris+.
Alan Smith, Ashutosh Jadhav , Hemant Purohit, Lu Chen, Michael Cooney, Pavan Kapanipathi, Pramod Anatharam, Wenbo Wang (Past Members: Karthik Gomadam, Meena Nagarajan)
Twitris- a System for Collective Social Intelligence
Amit Sheth, Ashutosh Jadhav, Pavan Kapanipathi, Chen Lu, Hemant Purohit, Gary Alan Smith, Wenbo Wang, Encyclopedia of Social Network Analysis and Mining (ESNAM).
Are Twitter Users Equal in Predicting Elections? A Study of User Groups in Predicting 2012 U.S. Republican Presidential Primaries
Lu Chen, Wenbo Wang and Amit P. Sheth, In Proceedings of the Fourth International Conference on Social Informatics (SocInfo'12), 2012.
Harnessing Twitter 'Big Data' for Automatic Emotion Identification
Wenbo Wang, Lu Chen, Krishnaprasad Thirunarayan and Amit P. Sheth, In Proceedings of International Conference on Social Computing (SocialCom), 2012.
Topical Anomaly Detection from Twitter Stream
Pramod Anantharam, Krishnaprasad Thirunarayan, and Amit Sheth, Research Note: In the Proceedings of ACM Web Science 2012, Evanston, Illinois, June 22-24, 2012.
What kind of #communication is Twitter? A psycholinguistic perspective on communication in Twitter for the purpose of emergency coordination
H. Purohit, A. Hampton, V. Shalin, A. Sheth, J. Flach, NSF SoCS Symposium, 2012.
Extracting Diverse Sentiment Expressions with Target-dependent Polarity from Twitter
Lu Chen, Wenbo Wang, Meenakshi Nagarajan, Shaojun Wang and Amit P. Sheth, In Proceedings of the 6th International AAAI Conference on Weblogs and Social Media (ICWSM), 2012.
Twitris+: Social Media Analytics Platform for Effective Coordination
A. Smith, A. Sheth, A. Jadhav, H. Purohit, L. Chen, M. Cooney, P. Kapanipathi, P. Anantharam, P. Koneru and W. Wang, NSF SoCS Symposium, 2012.
Discovering Fine-grained Sentiment in Suicide Notes
Wenbo Wang, Lu Chen, Ming Tan, Shaojun Wang, Amit P. Sheth, Biomedical Informatics Insights, vol. 5 (Suppl. 1) pp. 137-145, 2012.
Prediction of Topic Volume on Twitter
Yiye Ruan, Hemant Purohit, Dave Fuhry, Srini Parthasarthy, Amit Sheth, 4th Int'l ACM Conference of Web Science (WebSci), 2012.
Framework for the Analysis of Coordination in Crisis Response
H. Purohit, A. Hampton, V. Shalin, A. Sheth, J. Flach, Workshop in conjunction with CSCW-2012.
Personalized Filtering of the Twitter Stream
Pavan Kapanipathi, Fabrizio Orlandi, Amit Sheth, Alexandre Passant, 2nd workshop on Semantic Personalized Information Management at ISWC 2011.
Citizen Sensing - Mining Social Signals & Perceptions: Microsoft Research Faculty Summit
Amit Sheth, Invited Talk at Microsoft Research Faculty Summit 2011, Redmond, WA, July 19, 2011.
Understanding User-Community Engagement by Multi-faceted Features: A Case Study on Twitter
H. Purohit, Y. Ruan, A. Joshi, S. Parthasarathy, A. Sheth, Workshop on Social Media Engagement, in conjunction with WWW 2011.
Citizen Sensor Data Mining, Social Media Analytics and Development Centric Web Applications
Meenakshi Nagarajan,Amit Sheth,Selvam Velmurugan, Proc of the WWW 2011, March 28 - April 1, 2011, Hyderabad, India, ACM.
Twarql: Tapping into the Wisdom of the Crowd
P. Mendes, P. Kapanipathi, and A. Passant, Triplification Challenge 2010 at 6th International Conference on Semantic Systems (I-SEMANTICS), Graz, Austria, 1-3 September 2010. (Winner of Triplification Challenge 2010).
Linked Open Social Signals
Mendes PN, Passant A, Kapanipathi P, Sheth AP, WI2010 IEEE/WIC/ACM International Conference on Web Intelligence (WI-10), Toronto, Canada, Aug. 31 to Sep. 3, 2010.
Understanding User-Generated Content on Social Media
Meenakshi Nagarajan, Understanding User-Generated Content on Social Media, Ph.D. Dissertation, Wright State University, 2010.
Multimodal Social Intelligence in a Real-Time Dashboard System
Daniel Gruhl, Meenakshi Nagarajan, Jan Pieper, Christine Robson, Amit Sheth, VLDB Journal on 'Data Management and Mining for Social Networks and Social Media', 6 (2) 2010.
Twitris 2.0 : Semantically Empowered System for Understanding Perceptions From Social Data
A. Jadhav, H. Purohit, P. Kapanipathi, P. Ananthram, A. Ranabahu, V. Nguyen, P. Mendes, A. G. Smith, M. Cooney, A. Sheth, ISWC 2010 Semantic Web Application Challenge.
A Qualitative Examination of Topical Tweet and Retweet Practices
Meenakshi Nagarajan, Hemant Purohit, Amit Sheth, 4th Int'l AAAI Conference on Weblogs and Social Media, ICWSM 2010, pp. 295-298.
Some Trust Issues in Social Networks and Sensor Networks
Krishnaprasad Thirunarayan, Pramod Anantharam, Cory Henson, Amit Sheth, Proceedings of 2010 International Symposium on Collaborative Technologies and Systems (CTS 2010), Chicago, IL, May 17-21, 2010.
Understanding Events Through Analysis Of Social Media
Amit Sheth, Hemant Purohit, Ashutosh Jadhav, Pavan Kapanipathi and Lu Chen, Technical Report, Kno.e.sis Center, 2010.
Spatio-Temporal-Thematic Analysis of Citizen-Sensor Data - Challenges and Experiences
Meenakshi Nagarajan, Karthik Gomadam, Amit Sheth, Ajith Ranabahu, Raghava Mutharaju and Ashutosh Jadhav, Tenth International Conference on Web Information Systems Engineering, October 5-7, 2009, 539 - 553.
Citizen Sensing, Social Signals, and Enriching Human Experience
A. Sheth, IEEE Internet Computing, July/August 2009, pp. 80-85.
Analysis and Monetization of Social Data
Amit Sheth, Panel on 'Semantifying Social Networks,' Semantic Technology Conference, June 16, 2009, San Jose, CA.
Semantic Integration of Citizen Sensor Data and Multilevel Sensing: A Comprehensive Path Towards Event Monitoring and Situational Awareness
Amit Sheth, From E-Gov to Connected Governance: the Role of Cloud Computing, Web 2.0 and Web 3.0 Semantic Technologies, Fall Church, VA, February 17, 2009.
Picking the President: Twindex, Twitris Track Social Media Electorate
Web App Analyzes Tweets in Real Time for a Record of Historic Events
Twitris Social Media Analysis Tackles Occupy Wall Street, 2012 Elections
Wright State research seeks sense from social media to aid in emergencies
Citizen Sensor Data Mining, Social Media Analytics and Development Centric Web Applications