Early work on sentiment analysis focused on simple efforts to quantify sentiment as positive or negative. Today, sentiment analysis efforts today have gone much further. This talk presents a brief overview of recent trends in the field and will address progress in the following directions. How can we identify the emotional signals (e.g. anger, happiness, fear) in plain or noisy text? Can we use such signals to address business needs in multiple sectors? Can we quantify and predict the diffusion of sentiment on a given topic through online social networks? How would companies leverage this knowledge to solve business needs? The talk will include a lively set of real-world examples.

Social media PR crises make headlines frequently, but do the tidal waves of negative sentiment hurt brands? And many brands are accelerating their engagement strategies in social networks, but does all of this positive social sentiment build brands? We will explore where engagement and sentiment matters, and how companies must build brand vectors into their social strategies.

Sentiment Analysis of financial news is at this point a fairly well established field, though the interpretations of the problem differ, as do the approaches. It is however only a part of the solution to the overall problem of providing actionable analytics to decision makers in the financial markets. We will present the evolution of news analytics from Bloomberg's point of view and discuss possible future directions for the field (approaches to multilingual sentiment analysis and other indicators, relationship graphs analysis, etc.)

In March 2012, the CDC launched its first antismoking campaign called "Tips from Former Smokers". The campaign was regarded as controversial for using strong negative emotion and graphic images to promote cessation. Over 77,000 tweets and metadata were collected using Twitter's full fire hose, rather than the publicly available data stream. This allowed a more diverse collection of tweets and provided more robust metadata. The team reported on the overall reach and audience engagement of the campaign through an analysis of unique users reached, number of retweets, and mentions. Sentiment analysis was conducted to gauge the emotion of fear in tweets about individual television ads. Researchers at the University of Illinois Chicago's Health Media Collaboratory found that the effect of graphic images was broad and deep across tweets. The campaign achieved substantial engagement, rather than rejection of the promotion message.

I will share PayPal's approach, using basic sentiment/text analysis combined with operational excellence/six sigma principles, that resulted in X million of successful logins that produced revenue lift at PayPal and reduced call volume and costs.

A 2011 survey conducted by research firm StrategyOne found that family and social networks' role in personal health decisions is paramount. Thus as a natural extension, it would seem that listening online to identify influencers and leverage them to promote health interventions would be a critical aspect of public health communication. However, this has not been the case... until now. At Westat, we have been working with various government clients such as the Centers for Disease Control & Prevention (CDC), National Institutes of Health (NIH), and the Agency for Healthcare Research & Quality (AHRQ) to mine these data to find those influencers and impact health behavior. Whether focused on finding individual influencers to engage around health & healthcare issues, or taking a more B2B approach and identifying organizational influencers that reach our target audiences, we are listening to what is happening in this space and leveraging these data to drive our programmatic recommendations and results. This presentation will look at what tools, i.e. Radian6, Topsy, are being used; how government clients are using these tools; and finally what results have been captured, i.e. influencers identified, leveraging these for campaigns, and the impact.

When people communicate online, they use their knowledge of the world to talk creatively about everything from products to news. Gaining meaning from this text goes beyond keywords to involve a deep understanding of most of the explicit and implicit ways of communicating and talking creatively. Non-concept level approaches often rely on parts of text in which opinions and sentiments are explicitly expressed such as polarity terms, affect words, and their co-occurrence frequencies. However, this surface-level understanding often misses meaning. We explain how concept-based sentiment analysis solves this problem by using models of how people think about the world to reason about language without the use of rules, lexicons, training, or ontologies. We also show how we've duplicated this technique in nearly every written language -- without translation -- and linked these to provide cross and multi-linguistic reasoning that picks up nuance in real-world language.

Which candidate responses in a presidential debate did the audience perceive as dodging the question? What specific moments in a new comedy did TV viewers find hilarious, moving, boring, or painful? Questions like these are hard to answer. Traditional polling provides interpretable data, but it can't give you second-by-second information about people's reactions, and it doesn't scale. Response dials allow fine-grained temporal feedback, but on just one dimension. Social media analysis gets nearer to real time, and on a massive scale, but it doesn't allow researchers to pose specific questions to the audience.

This presentation will discuss React Labs, a new "best of breed" system that allows viewers of an event to react in real time on mobile devices to events they are watching, combining elements of discrete input and social media experience in order to create user engagement and collect temporally fine-grained data. Originally inspired by the goal of identifying "spin" in political debates, the framework is applicable in a broad set of use cases -- a set that is growing rapidly with increasingly widespread tablet and smartphone use while watching television and live events.

The recent explosion of "second screen" experiences, services, and apps (which has been frequently coined "SocialTV") has always been a temporary phase to fill in the disconnect between the Networks, Advertisers, and the Viewers. As the usage of those services and apps has begun to finally transition from opt-in to an expected integration, we are witnessing Networks and their Advertisers rebuilding their personal relationships with their viewers. The difference this time though is that social allows this to be an "always-on" relationship. These "always-on" relationships have created the ability for Networks to leverage the ever-growing data created by those Viewers to produce programming that resonates and guarantees an engaged audience before the show actually airs. Leveraging viewers in this fashion allows Advertisers to feel confident in advertising on a new program and informs how to create content that appeals to that audience, thus extending their reach beyond a single TV show. Real-time marketing solutions like Networked Insights, empower Networks and advertisers to achieve this ?always-on? relationship. Sean Reckwerdt, Lead Analyst at Networked Insights, will discuss how technological advances in social media analytics, machine-learning organic discovery, and the way brands are marketing have allowed this revolution to occur.

We use market research -- including text and sentiment analysis -- to measure customer loyalty to ensure the company to focus attention on the 'right' customers, improve team efficiency, create customer advocates, and create a baseline for measuring change. We

• Conduct 'Moments of Truth' sentiment analysis to identify and close gaps between how customers and employees view customer centricity
• Hold everyone accountable for moving the needle and closing the performance gap
• Relate the number of net promoter recommendations back to the bottom line

What is new about this approach is that we use customer research to develop key drivers of the business based on sentiment analysis and then conduct employee research to determine if employees and customers are in alignment in regards to whether the company is delivering an extraordinary experience. This research is then used to develop strategic action plans which are monitored on a quarterly basis through on-going research studies.

A recent survey by Opinion Research Corp found that online customer reviews influenced the purchase decision of a product or service for 84% of Americans. Similarly, a Bernard Hodes Group study found that 54% of people were influenced to consider working for an employer. Views of online reviews and forum websites have grown exponentially - websites like Reddit, Glassdoor, and Amazon contain thousands of easily findable comments that affect brand reputation and employer identity. In creating a method of tabulation, measurement, and analysis, we have developed a process to understand current trends regarding Brand identity. Though manual, utilizing 'Topic Trends' found in review sites and forums to inspire content strategy focuses our content and produces consistent quality content that resonates and engages. Over time, changes in sentiment and topics regarding trends gives the strategist unique opportunities. This process translates well for Brand and Product identity and reputation management.