Agenda
The program for the March 5, 2014 "human analytics" workshops and the March 6, 2014 conference is outlined in the agenda that follows. Attend either or both. And whichever day(s) you attend, join us for the Wednesday evening, March 5 networking reception, starting at 5:30 pm.
March 5 Workshops | Business Track | Technical Track | |
Morning Session | |||
9:00 am-12:30 pm |
Stephen D. Rappaport, Stephen D. Rappaport Consulting LLC
with Vincent Santino, Kaplan Test Prep; Peter Fontana, We Are Social; and Maribel Lara, M80
|
Jason Baldridge, University of Texas
|
|
Afternoon Session | |||
1:30 pm-5:00 pm |
Steve Ramirez, Beyond the Arc
|
technology oriented presentations
|
|
Networking Reception | |||
5:00 pm-6:30 pm | Networking Reception Join us for drinks and hors d'oeuvres |
March 6 Conference | |||
Morning Session | |||
8:00 am-8:30 am | Registration & Coffee | ||
8:30 am-8:40 am | Chair's Welcome: The Sentiment Spectrum
Seth Grimes, Alta Plana Corporation
|
V I S I O N & A P P L I C A T I O N |
|
8:40 am-9:15 am | Measure and Predict: Lessons in Customer Satisfaction and Loyalty
Bob Hayes, TCElab
(mouseover here for description)
Customer satisfaction and loyalty are important variables in today's
business world. Senior executives rely on measures of these variables to
gain deeper insights about their customers. Businesses have access to
droves of different types of measures that tap both satisfaction and
loyalty. Applying predictive analytics to these measures, executives try to
improve how they market to, sell to and service their customers.
In this presentation, customer experience visionary, Bob Hayes, will provide an overview of the different types of Voice of the Customer data (behaviors, transactions, attitudes, outcomes, etc.) and analytic methods used to draw these insights. He will show how businesses can gain insight from both structured ratings and text-extracted information that can help businesses improve both strategic and tactical decision making. |
||
9:15 am-9:45 am | Adventures in Emotion Recognition
Rosalind Picard, MIT Media Lab
(mouseover here for description)
At the MIT Media Lab, we have been creating technology to recognize and better communicate emotion in the real world where emotions tend to signal "what really matters". This talk will show
examples and reveal surprises we encountered when doing this work. Examples include creating the world's first automated facial affect recognition online, which now has over a million
facial expression videos from people opting in to express their feelings by having the camera watch them while they watch online media. I will also highlight some surprises in training
machine learning to recognize states such as frustration: While most people have difficulty discriminating "smiles of frustration" from "smiles of delight" in static images, we are now
able to get the computer to be highly accurate at discriminating these. These new affect-recognition tools can potentially help people with nonverbal learning disabilities, limited vision,
social phobia, or autism who find it challenging to read the faces of those around them. I will also share recent findings from people wearing physiological sensors 24/7, and how we've been
learning about connections between the emotion system, engagement, sleep and seizures. Finally, I will share some of our newest work related to crowd sourcing cognitive-behavioral therapy
and computational empathy, where sentiment analysis and topic modeling are of benefit.
|
||
9:45 am-10:10 am | Mapping Human Motivations to Move Product
David Rabjohns, MotiveQuest
(mouseover here for description)
There are three levels at which we
can explain customer behavior: 1) Features; 2) Benefits sought; and 3) Underlying Motives. Customer motivations are
universal; a common set of motivations collectively paints a rich picture of the drivers of customer behavior.
Measure and map customer motivations based on Online Anthropology for a powerful tool that can be used to:
|
||
10:10 am-10:35 am | Break | ||
10:35 am-11:15 am |
Visionaries Panel
Stephen D. Rappaport, Stephen D. Rappaport Consulting LLC, moderator
|
||
11:15 am-11:40 am |
Social Intelligence at the Crossroads
Chris Boudreaux, Accenture
|
||
11:40 am-11:50 am |
Strategies for Successful Sentiment Analysis of Realtime Social Data
Scott Hendrickson, Gnip
|
||
11:50 am-12:15 pm |
Lightning Talks
5-minute presentations/demos of human analytics technologies, solutions, and services
|
||
12:15 pm-12:30 pm |
Accuracy and Opportunity
Seth Redmore, Lexalytics
|
||
Lunch & Networking | |||
12:30 pm-1:30 pm | Lunch & Networking | ||
Afternoon Session | |||
1:30 pm-2:05 pm | Engagement: The Unspoken Connection
Marie Wallace, IBM Corporation
(mouseover here for description)
As the saying goes, "Actions speak louder than words." In this session we
will examine the role that engagement analytics can play -- how people act and interact and what meaning those connections communicate -- in helping us
truly understand what is happening inside collaboration & social networks.
Are people happy or apathetic? How is information flowing through the network? Who influences? Who is influenced? Who are the information hubs? Who is on the periphery? Evaluating the pattern of interactions inside the network can frequently tell more than the content, or at least can put the content into context. We will also describe a number of scenarios where engagement analytics can drive business value and help organizations become more people centric in everything they do and every decision they make. |
E X P E R T I S E |
|
2:05 pm-2:25 pm |
Forecasting the Spread of Sentiments and Emotion in Social Media
VS Subrahmanian, Univ of Maryland and SentiMetrix
(mouseover here for description)
There are numerous applications where we need to forecast the spread of sentiment and emotion across social media. Recently, Facebook had to assess the number of people (and importance of
people) angry at their policy to permit the posting of beheadings on their network. Other companies have had to roll back pricing policies and/or fees in the face of a swell of criticism
on the Internet. Politicians need to understand emotions of the voting public in order to best craft a message that will get them elected – for example, they may wish to forecast the
intensity and extent of anger at issues like corruption and the intensity and extent of fear of issues such as unemployment. In this talk, we will use a major forthcoming international
election as a colorful case study of how to track sentiment and emotion on any given topic, make forecasts about how far those sentiments and emotions will spread, and suggest ways in
which to maximally influence the outcome to be consistent with the needs of client.
|
||
2:25 pm-2:45 pm | Extracting Trading Signals from Investor Sentiment
Sarah Biller, Capital Market Exchange
(mouseover here for description)
After an eye-opening event like the credit crisis, why have
things not changed in a professional bond investors workflow? They use the same models and the same backward looking data and trade just as they have since the 1940s, only to find that
the insights they seek allude them.
And, to overcome this lack of market information, asset managers spend over $25 billion annually on data and analytics. Where the bond market structurally fails them, actionable sentiment analysis solves the void. I will share Capital Market Exchange's approach to aggregating and analyzing current market sentiment to provide clients at firms like Oppenheimer and Citigroup actionable outcomes, to provide valuable lessons for all who seek to act on signals suggested by sentiment. |
||
2:45 pm-3:05 pm |
Sentiment Analysis for Comment Moderation
Maggie Xiong, The Huffington Post
(mouseover here for description)
At The Huffington Post, one of our top priorities is engaging our readers' active involvement by encouraging their participation in comment conversations around the news we
report. A key objective here is to maintain an atmosphere of civility in these conversations, and above all to ensure they are not hijacked by "trolls" -- bad actors whose purpose is to
disrupt, disparage, and ultimately derail otherwise fruitful discussions.
To do so, HuffPost employs a team of forty moderators, all but one of whom are flesh-and-blood humans. The exception is a state-of-the-art machine learning system called JuLiA (for "Just a Linguistic Algorithm'), which employs sentiment analytics to shoulder the burden of moderating some three quarters of the 250,000 comments we receive on an average news day. As capable as JuLiA is, we've recently embarked on an effort to make 'her' even better. We?re doing this by leveraging the judgments of our human moderators to continually amass up-to-date data for use as a "gold standard" against which to train an ongoing series of new sentiment-analysis models. These new models will not just trap generically abusive comments, but will also focus in on particular topic areas where reader emotions tend to run high, and ensure the related discussion threads continue to generate more light than heat. Further out, we plan on using our patented sentiment-tracking Huffington Harmony Index (HHI) to zero in on those articles which are generating high proportions of abusive comments and build special-purpose models on the fly to counter them. |
||
3:05 pm-3:15 pm |
Deriving Social ROI Through Enhanced Text Analytics in Real Time
Shree Dandekar, Dell Software
|
||
3:15 pm-3:35 pm | Break | ||
3:35 pm-4:10 pm |
Innovation Panel
Steve Ramirez, Beyond the Arc, moderator
|
I N N O V A T I O N |
|
4:10 pm-4:30 pm | Bleeding Edge NLP: Extract & Analyze Events, Relationships, Emotion, Intent, Identity, and Lots of Other Good Stuff
Stephen Pulman, University of Oxford
(mouseover here for description)
Finding named entities in text is often regarded as a done deal, whereas in
fact in many applications this is
still a challenging task. Still more so is the extraction of events and
relations, emotion, intent, identifying entities, classifying authors
by gender, age, etc. This talk will give an overview of recent research in
these areas, covering the use of 'deep learning' and other
currently hot topics.
|
||
4:30 pm-4:50 pm | "Let's Get Emotional": Emotions in Speech
Dan Emodi, BeyondVerbal
(mouseover here for description)
Emotions Analytics is a new field that focuses on identifying and analyzing the full spectrum of human emotions and personality out of people raw vocal intonations.
Introducing emotions understanding into mobile applications, enterprise solutions and devices opens up a new dimension of person-machine interface and finally enable machines to understand who we are, how we feel and what we really mean. This session will describe why understanding emotions is one of the most important up-and-coming interfaces and will describe the best practices of implementing Emotions in voice powered and voice enabled solutions.
|
||
4:50 pm-5:10 pm | Automatic Facial Expression Recognition for Emotion-Aware Computing
Jacob Whitehill, Emotient
(mouseover here for description)
As natural user interfaces evolve, we believe the next phase will
harness technology to detect users' emotions, level
of interest and engagement, and even predict intent to purchase. In this
presentation Dr. Jacob Whitehill will (1) explain a state-of-the-art approach
that applies machine learning, computer vision and cognitive science to
facial expression recognition; (2) demonstrate various emotion outputs; and
(3) highlight market opportunities to leverage an emotion recognition system.
|
||
5:10 pm-5:30 pm | A New Approach to Real Time Intent and Sentiment Analysis
Aloke Guha, Cruxly & Hitachi Data Systems, and Kapil Tundwal, Cruxly
(mouseover here for description)
Ability to detect actionable intent in online and messaging text, especially in near real-time, is becoming significant in customer acquisition, marketing, support and product
management. Location based services and wide-spread adoption of mobile devices further increase the importance of detecting intent to buy, a customer commitment, and event occurrences.
In this talk, we will present an approach that applies NLP techniques for intent detection in real-time. The approach is based on a semi-supervised technique that combines grammar rules with shallow parsing. Intent to Buy, Questions or Requests, Like, Dislike, Commit and Recommendations: Combinations of these intents can be used in areas from lead generation to customer support, even in noisy posts such as in Twitter. |
||
5:30 pm |
Conference Close: A Look Ahead
|