About Us
Traditionally, public opinion has been measured by polling, surveying and focus groups. The explosion of social media in recent years has created a new and large online data set that can be used to understand public opinion and measure reaction. The Social Reaction Group is leveraging this new data to study public opinion, and in particular, examining how public opinion is affected by demographics, psychographics, location, trends, and other variables.
To do this work, we are using technology developed at InferLink Corporation. The infographics on this website are relatively simple demonstrations of what’s possible, they show how Twitter data about public speeches can be analyzed. The same approach can be applied to a variety of public opinion and reaction analysis in both the public and private sector. If you have ideas for new applications, want to find out about our work, or are interested in having us conduct a study, please contact us.
(Thanks to SDL for providing the text translations on this site).
Methodology
(Note: This section is written for the general public, and is not a complete description of the technical process used to analyze speeches. Contact us for technical details.)
Step 1: Collection
During collection, Social Reaction group uses data aggregation techniques to collect as much relevant data from Twitter as possible. This includes gathering tweets from anyone who is discussing the speech or topics around the speech.
Step 2: Demographic Identification and Categorization
The tweets are analyzed and categorized into a number of different groups based on demographic and psychographic variables, such as gender or location.
Step 3: Topic Clustering
During this phase, the collection of tweets are analyzed to detect topics of discussion, and each tweet is categorized accordingly. For example, if a tweet was about Michelle Obama’s dress, that tweet would be placed in a group with other tweets commenting on her dress.
Step 4: Topic to Sentence Alignment
Next the topic clusters are mapped to sentences in the speech. This allows us to look at social reaction within the speech on a line by line basis. We also identify topics that spike during certain parts of the speech, but do not directly align with a particular sentence. These are called “events”.
Step 5: Statistics
Once we have mapped all of the tweets to sentences, we begin to see patterns in the data. We then run statistical tools to determine the significance of the patterns. For example, we might look to determine whether a higher volume about a particular sentence indicates statistically significant information.
Members of the Social Reaction Group
Dr. Steven Minton is the President of InferLink Corporation. He is an expert in AI and a fellow of the AAA. He previously founded Fetch Technologies and the Journal of AI Research.
Dr. Sofus Macskassy is a member of the faculty of University of Southern California’s Computer Science Department, and a Project Leader at the Information Sciences Institute. He is an expert in artificial intelligence, machine learning, text mining, and social network analysis.
Dr. Shibley Telhami is the Anwar Sadat Professor for Peace and Development at the University of Maryland, College Park, and non-resident senior fellow at the Saban Center at the Brookings Institution. He conducts extensive public opinion surveys in the Arab world. He is a member of the Council on Foreign Relations and served on the board of Human Rights Watch (and as Chair of Human Rights Watch/Middle East).
Karl Stark is the assistant managing editor of business coverage and health and science coverage at the Philadelphia Inquirer. He also writes weekly jazz reviews for the newspaper.
Dr. Greg Barish is the Chief Technology Officer at InferLink Corporation. He has significant research and development experience combining artificial intelligence & parallel processing techniques for integrating and analyzing Web data.
Brian Amanatullah is the Director of Client Solutions at InferLink Corporation. He is an expert in developing enterprise software systems. He did most of the real work to create the infographics on this site.