Agenda

The Sentiment Analysis Symposium program for Tuesday, April 12, 2011 is laid out in the agenda that follows. Click here for the agenda at a glance.

Monday afternoon, April 11 2:00 pm-5:30 pm	Pre-symposium tutorial: Sentiment Analysis in Practice (optional, separate registration fee) Yongzheng Zhang, Catherine Baudin, and Nitin Indurkhya, eBay Research Labs Click here for the tutorial description.
Monday evening, April 11 5:30 pm-7:00 pm	Pre-symposium get-together for attendees and friends at the Glass House Tavern, 252 W. 47th Street.
Tuesday morning, April 12 8:00 am-9:00 am	Check-in & Coffee
Morning Session
9:00 am-9:15 am	Chair's Welcome Seth Grimes, Alta Plana Corporation
9:15 am-10:05 am	Visionaries Panel The Visionaries Panel delves into the business impact of extracting value from opinions and attitudes in social media, news, and enterprise feedback. Moderator: Steve Rappaport, Knowledge Solutions Director, Advertising Research Foundation Jeff Catlin, CEO, Lexalytics Israel Mirsky, EVP, Emerging Media & Technology, Porter Novelli Katie Delahaye Paine, CEO, KDPaine & Partners Karla Wachter, SVP, Waggener Edstrom
10:05 am-10:40 am	Sentiment Analysis Past, Present & Future, and an Application in Trading Strategies Ronen Feldman, Associate Professor, Hebrew University and Co-founder and Chief Scientist, Digital Trowel
10:40 am-11:15 am	Mining Community Voices: Topics, Trends, Emotion and Opinions in eBay Community Data In a company like eBay, community discussion boards are sources of information ranging from user reactions about new search /UI features to general opinions about products. Discovering topics and sentiment from opinion pages and classifying them provide insights into user's issues and experience, and as such is a valuable complement to user studies and formal product reviews. Accurately mining such unstructured and conversational data however, is challenging. In particular, users post for different purposes: socializing, requesting information, reaction to events, or product discussions, and these different motivations are often associated with different levels of emotion. For example opinions about a new digital camera are likely to be expressed differently than sellers' reactions to a change in the search engine that may impact their livelihood, which may, for example be associated with high levels of sarcasm. We are reflecting on our experience with different topic and opinion extraction prototypes, including Vox Populi, a tool to help extract topics and classify user sentiment in eBay forums and in unstructured product reviews. Catherine Baudin, Senior Research Scientist, eBay Research Labs
11:15 am-11:30 am	Break
11:30 am-12:30 pm	Lightning Talks (11 slots) A series of quick presentations/demonstrations of sentiment-analysis and related solutions. Bryan Bell, VP, Enterprise Solutions, Expert System Fiona McNeill, Global Product Marketing - SAS Text Analytics, SAS Steve Pettigrew, Technical Product Manager, Content Analytics, OpenText Karo Moilanen, Oxford University Seth Earley, President & CEO, Earley & Associates Marguerite Leenhardt, Paris 3 University Lawrence C. Rafsky Ph.D., CEO, Acquire Media Lipika Dey, Senior Consultant, Tata Consultancy Services Taras Zagibalov, Brandwatch Shane Axtell, Computational Linguist, Clarabridge Elliot Bricker, Director, Product Management, NetBase Solutions, Inc.
Lunch & Networking
12:30 pm-1:30 pm	Lunch & Networking
12:30 pm-5:40 pm	Exhibit Open
Afternoon Session
1:30 pm-1:50 pm	Lightning Talks: Round 2 (3 slots) Pete Snell, President, Peter Snell & Associates Rich Brown, Global Business Manager, Machine Readable News, Thomson Reuters Max Yankelevich, Chief Cloud Architect, Freedom OSS, and John Hoskins, Amazon.com
1:50 pm-2:25 pm	Understanding the Voice of the Customer in Mobile Devices: Integrating Survey and Social Media Research Companies worldwide invest over $1 billion annually measuring the customer experience. Yet, even with the enormous data volume available and sizeable resource expenditures, many companies still face significant challenges in creating a comprehensive view of customer feedback. Furthermore, companies garner praise and criticism from customers -- and non-customers -- online via review websites, Twitter, blogs and other consumer-generated media. Millions of people use this chatter to make purchase decisions. Indeed, in the US, 24% of adults have written a product review online and 58% conducted research online about products and services they buy. Online commentary is too important to be ignored. Thus companies must integrate multiple sources of voice-of-the-customer (VOC) data as a way to better leverage existing investments in primary research and make sense of the social media landscape in order to improve action planning. Through a multi-brand study of consumers who purchased or influenced the purchase of a Smartphone or Tablet device during the holiday shopping period, this presentation showcases the integration of survey research (provided by the Mobile Product Launch Monitor survey conducted by Maritz Research) and social media research (provided by evolve24, a Maritz Research company) to examine consumer sentiment during the holiday shopping period about their purchase and product experiences, including handset, applications, and carriers. Text mining, sentiment analysis, and data mining software have roles in the analysis. Manufacturers and carriers can use the integrated results to guide product design, marketing, marketing communications, and product management. Michael House, Division VP, Maritz Research
2:25 pm-3:00 pm	Computers That Trade Off The News - Real Applications Factoring news, an inherently qualitative source of information, into market models requires the fast and accurate conversion of qualitative information into reliable and consistent numerical data streams. The volume of news alone presents challenges that until recently have had only theoretical solutions. For example, relevant content for publicly traded companies across major global indexes, from the combination of the professional newswires, consumer media, and top web publishers exceeds one hundred thousand original stories a day - something impossible to digest by humans. This presentation will provide insights on how financial institutions are actually using computers to read and interpret news. Computers are now being used to analyze news information to aid in the valuation and trading of securities, facilitate investment decision making, meet regulatory requirements, and manage risk. Users of these new technologies include some of the top financial institutions in the world including the best performing hedge funds, asset management firms, brokerages, and global banks. Beyond discussing the more known high-frequency trading applications, the presentation will cover how groups running "quantitative investment strategies" are finding how news and sentiment can be a powerful stock selection and risk factor -- and particularly how news sentiment is uncorrelated with traditional factors, adding diversification benefits to investment models with short or long term horizons, and high or low frequency trading applications. Finally, the presentation will disclose how news analytics have worked at times when many traditional factors have failed. Armando Gonzalez, CEO, RavenPack
3:00 pm-3:35 pm	Movie Sentiment Ranking Using Twitter Data Aol has developed a set of sentiment analysis algorithms that uses the Twitter stream (or any other UGC stream) to rank movies according to the sentiment users express towards them. The movie ranking system generates an up-to-date picture of what the users think about newly screened movies and can be compared to the view critics hold towards these movies. By monitoring the behavior of upcoming movies the system can extract the "hype" level around a soon-to-be screened movie and can be a relatively good predictor to the success of a movie in the box-offices on its first weekend. The system identifies (in near-real-time) those tweets where movies (currently new and upcoming) are being mentioned and uses state-of-the-art sentiment analysis to identify positive or negative attributes towards the movie. By using a mixture of NLP pattern analysis and supporting rule the system extracts the sentiment for each relevant tweet and assigns a confidence level to the finding. The system then aggregate the results for each movie assigning an overall sentiment score for each movie. The final results is a rank on a positive-negative scale for each movie. Besides the current ranking, the system also monitors the sentiment volume trends towards the movies and can identify "hot" movies, ones that experience a surge in their sentiment volume. The algorithms used circumnavigates many of the pitfalls shared by similar features by applying several statistical models when analyzing the aggregated results. Amit Moran, R&D Manager, Aol
3:35 pm-3:55 pm	Break
3:55 pm-4:30 pm	Beyond Sentiment: How to Really Get Value out of an Automated System Too much attention has been focused on sentiment as the qualtitative factor in social media measurement. In fact, there are many other, and arguably more accurate, criteria that automation can tell you about what the market is saying about you. This presentation provides 7 basic steps to getting what you really need out of your sentiment analysis system. Katie Delahaye Paine, CEO, KDPaine & Partners
4:30 pm-5:05 pm	The Right Formula for Sentiment Success: Emotion, Clarity, and Intensity Many have criticized Sentiment Analysis, mostly due to the fact that merely labeling text positive, negative or neutral provided an enormous opportunity for error. Words, like emotions, aren't black and white and there are several other factors to be considered before an analysis can be considered "accurate." Considering decision-making is predominantly fueled by emotion, isn?t it critical we get it right? Josh Merchant, CTO at Lymbix Inc, will shed some light on the subject of accuracy and why it takes more than a positive/negative indicator to provide a true analysis. In addition to providing insight on three key sentiment layers (specific emotions, clarity, and intensity), he looks forward to sharing some of his own early struggles finding the "right formula" for success in the sentiment space. Josh Merchant, CTO & Co-Founder, Lymbix Inc
5:05 pm-5:40 pm	Milan Goes Social: Customer Experience, Reputation, and Semantics for Tourism Milan hosts about 4 million visitors every year, with an 18% growth over the past ten years. However, it has been estimated that business tourists account for over 70% of total guests. The goal of the project was to understand how social media can help reposition Milan in the short-break market, in order to increase leisure tourism. Social media were seen as an unbiased source of information to understand the current strengths and weaknesses of Milan as a tourism destination. As a first step, we have adopted Anholt's place branding model to describe a city's touristic attractiveness. We have defined the semantic model (categories and keywords) that allow a mapping between social conversations and our model of a city's touristic attractiveness based on a sample of city Web sites (supply) and a preliminary manual screening of online conversations (demand). Third, we have designed a new semantic network for the English language embedding two types of domain knowledge: 1) general domain knowledge, represented as a sentiment polarity attached to relationships between words and 2) contextual knowledge, such as domain-specific proper names. We have monitored Facebook, Lonely Planet, Trip Advisor, and Twitter starting from April 2010. We have compared Milan with Berlin, Madrid, and London based on over 15 million posts. Analyses show how Facebook and Twitter allow a general ("vibrant") characterization of each city's attractiveness over extended periods of time. However, a real-time interpretation of data requires knowledge of a city's people (local/guest influencers) and "happenings". Both levers can be used to reposition Milan in the travel social fan segment. We have conducted a few pilot tests to understand how to leverage business presence at Milan's international fairs of design, fashion, ICT, and travel to increase leisure tourism through crowd-sourcing intelligence. Chiara Francalanci, Professor, Politecnico di Milano
5:40 pm	Wrap-up
Reception
5:40 pm-7:00 pm	Networking Reception Sponsored by Lexalytics

Register Today!

Agenda

Register Today!

Our Sponsors

Platinum

Gold

Bronze

Media