Agenda Details

April 13th, 2010

8:00am -

Registration & Coffee


9:00am -

Chair’s Welcome

Speaker:
Seth Grimes, Alta Plana


9:30am -

Visionaries Panel

The panel, moderated by industry analyst, Suresh Vittal, from Forrester will delve into the business impact of extracting value from opinions and attitudes in social media, news, and enterprise feedback.

Moderator: Suresh Vittal, Vice President & Principal Analyst, Forrester Research

Speakers:
Bradley Honan, Senior Vice President, StrategyOne
Greg Radner, Global Head of PR Services, Thomson Reuters
Brad McCormick, Executive Vice President and Director of Digital, Porter Novelli, Porter Novelli
Karla Wachter, Senior Vice President, Waggener Edstrom Worldwide


10:30am -

What Business Innovators Need to Know about Sentiment Analysis

Speaker:
Claire Cardie, Cornell University


11:05am -

Break


11:20am -

Lightning Talks

Entity Level Sentiments - Jeff Catlin, CEO, Lexalytics

One of the trickier parts of providing text analytics in foreign languages like Spanish or Portuguese is providing the entity level sentiments that are absolutely necessary in Financial Services and Reputation Management arenas.  We've introduced a new training based technique for entity sentiment, that allows us to rapidly build entity sentiment into foreign language ports of our text analytics engine.  The first example of this is being done in Portuguese, but will also be used to provide this capability in French and Spanish in the coming months.

Dynamic Ontologies: The next wave of consumer sentiment analytics - Steve Alexander, SrVP & Ravi Condamoor, CEO, Serendio

Sentiment mining from all forms of qualitative, unstructured customer feedback data is evolving as a key source for insights on products and services. The fast growth of the social web along with the proliferation of customer-care emails, call center narratives etc. has forced companies to look for new ways to make all this data more actionable.  However, a simple keyword based, domain agnostic approach to analyzing such data makes it extremely difficult to extract nuanced insights essential for any action or decision. The problem is that the metadata, primarily a keyword set, is too coarse for thoroughly understanding the factors driving sentiment and opinion.

Sentiment Analysis for Customer Experience Management - Justin Langseth is President & CTO of Clarabridge, Inc.

A quick overview of how sentiment analysis is best applied to the world of Customer Experience Management.  How to  determine what people like and dislike about your products and services, those of your competitors.  Also, how to determine the root causes and impacts of those sentiments.

Sentiment trends - Mattias Tyrberg, CEO Saplo

How can we in real-time follow and react on real-life discussions, for example a TV debate? Looking at the real-time web it's important to understand and react on trends. One example of users is politicians that need to get feedback from their voters. In the future we may see politicians that actually change or at least clarify their statements thanks to feedback from micro blogs such as Twitter. How can politicians and companies use this for their benefit? Mattias Tyrberg will talk about the topic and DEMO an alfa version of an analysis tool where you can track and analyse sentiment trends.

Successfully Analyzing Sentiment - Fiona McNeill, Global Product Marketing Manager - Text Analytics, SAS Inc.

Machine learning methods used to understand sentiment, without applying the experience to refine rules can result in misinterpretations and false conclusions.  Human experience is required to understand the subtleties and nuances that can evade algorithms. Sentiment analysis is often approached as a closed system ­ failing to permit the ability to intervene in accounting for slang, sarcasm or other colloquialisms ­ common to understanding nuances of sentiment, particularly when analyzing social media.  A hybrid approach is required. In this session, learn how SAS is empowering organizations to simultaneously leverage the latest in advanced text analytics using a hybrid approach ­ - one that ensures the consistency provided by statistical rigor while also allowing for the direct human intervention necessary to ensure sound analysis.

MarkLogic Server: Delivering Smart Information Products - Seth Altman, Senior Consultant, Mark Logic Corporation

MarkLogic Server was designed specifically as a platform for building agile content applications that unlock the full value of your information. Leaders in information and media, government and other industries have all used MarkLogic Server as the platform for their mission critical, role- and task-aware content applications. Unlike relational database and search hybrid solutions, MarkLogic Server is uniquely able to deliver content applications, which meet a broader range of user requirements by fully leveraging contents' XML structure. That's why leading businesses and government organizations rely on Mark Logic to give them the agility to quickly adapt to changing market conditions and new product requirements such as leveraging output from sentiment analysis tools to launch smarter applications and information products.

Tracking the Launch of Motorola Droid / Milestone in Social Media - A Sentiment Analysis Case Study - Dr. Ole-Christoffer Granmo, Associate Professor University of Agder; Research Fellow, Integrasco

Integrasco is a company that provides brand tracking in social media, including monitoring of buzz, sentiment, and user engagement. In November 2009, Integrasco tracked the launch of the Motorola Droid / Milestone smart phone, based on automated analysis of millions of discussion forum-, Twitter-, Facebook-, and YouTube-entries. In this talk, we will explain the role machine learning based sentiment analysis played in obtaining crucial findings about the Droid / Milestone. Furthermore, we discuss challenges we faced with respect to applying machine learning based sentiment analysis in this domain, such as the massive amounts of data available, the lack of appropriate training data, the need for a domain dependent sentiment vocabulary, as well as the typical presence of different kinds of textual nuisances, such as spelling and grammatical errors, abbreviations and Internet slang. Finally, we indicate how we dealt with some of these challenges, allowing the resulting sentiment analysis scheme to be successfully applied in a number of intriguing brand tracking cases in the following months.

Twitter Sentiment - Alec Go, Student, Stanford

Twitter Sentiment is a sentiment analysis tool for Twitter. It can be found at http://twittersentiment.appspot.com/. We'll discuss the different use cases for using the site.

4 Building Blocks of Superior Sentiment Analysis - Brooke Aker, CEO, Expert Systems

We will show how sequential building blocks of sentiment analysis result in superior business oriented applications.
The first building block of superior sentiment analysis are robust crawlers and converters to acquire social media regardless of its placement, format or speed at which it is updated. The second block is Natural Language Processing (NLP) or sometimes called text mining. NLP is technology that can read and comprehend social media or other text based inputs for the sentiment analysis. The NLP must be robust, accurate and flexible. The third building block is a relational database and sets of connectors that can take the output of the NLP building block and fill database. The fourth building block is clean and clear graphical analytics in the User Interface. The UI must be training-free, intuitive and a powerful query tool to the relational database.
Taken together these building blocks form a sound and proven basis for superior sentiment analysis.

Next generation sentiment extraction: light at the end of the tunnel, but will it negate the need for human supervision? - Sally Church, EVP, Icarus Consultants

The subtleties of human language makes the determination of sentiment extremely difficult via standalone technology.  Solutions abound, but without the overhead of skilled human set-up and supervision the results can be very misleading or simply wrong.  And the risk of getting it wrong means that organisations lack the confidence to embrace widespread technology application. Current systems require input of typical sentiment 'seeds' or 'phrases' (emotive terms) to help steer the technology into identifying sentiment commentary.  But such lists: need up-front creation and require on-going data management; need focus on particular subject areas - sentiment terms for Food are different to Insurance; suffer from the ambiguity of language and context - does 'better' refer to a gambler in the on-line gaming world, or good-quality product/service, or the negative comment 'could do better!'; foster 'keyword search' approaches, where one also needs to consider context to determine true meaning. There is real light at the end of the tunnel with the release of Leximancer's Automatic Sentiment Lens. 

Demonstration: Tweetsentiments - Dave Naffis, Cofounder & Tom Zeng, Software Architect, Intridea, Inc.

Tweetsentiments is a web services product that analyzes tweet sentiments using Machine Learning and Natural Language Processing technologies.  We will show those features: tracking positive/negative sentiments on topics for brand and event monitoring; tracking user's sentiments over time based on tweets; sentiment index by country using thematic maps, and overlay with other maps (i.e. H1N1) to detect possible correlations; using tweetsentiments.com API to add sentiment analysis capablities to 3rd party apps; various sentiment related charts and reports.

Sentiment and Topic Discovery over Twitter--What you can build in an afternoon with LingPipe -Breck Baldwin, President, Lingpipe

This talk starts with the Twitter search engine and shows the steps of applying LingPipe's clustering and classification classes at the source code level. From a raw query I will use LDA (latent Dirichlet allocation) to identify interesting topics as a data exploration tool. Next we cover using a spread sheet to create training data for the Twitter search results and then apply a logistic regression classifier and evaluate results. A link to the source is available at http://alias-i.com/drop/sentenceSentiment.tgz

Exhaustive Extraction - Ian Hersey, CTO, Attensity

Unlike many other systems, our semantic engines "start with the text itself" and lets it tell the story. Attensity invented (and has patented) a completely unique method for getting to accurate sentiment - we call it Exhaustive Extraction. This technology enables us to parse sentences and accurately look for instances of sentiment expression and get it right more so than any other vendor that attempts to do this.  The process is analogous to what we learned in grammar school English classes. Next we extract and aggregate the facts from the content. This allows us to not only find the words that are indicative of sentiment, but to find the relationships between words so that we can accurately identify both words that modify the sentiment (even if they are not close to the sentiment) and what the sentiment is about. Not only do we pull out the facts - but we can also find the degree of sentiment not just "love" but "really love" and not just "bad" reception but "very bad."

Five reasons candidates will need sentiment analysis (in addition to polling) to win in November - Laurel Earhart, Vice President, Business Development, SentiMetrix

My colleague, Todd Herman, of the Republican National Committee said it best: “Polls are like watching animals in a zoo. You can obtain perfect specimens, observe them from every angle, but then you wonder why they aren’t mating. Sentiment Analysis is like watching those animals in their natural habitat.” I think we all have a lot to learn from this statement. Polls are important, but as you’re developing a social media strategy, consider what sentiment analysis can do for you to win in November.

Understanding your market's dynamics with Sentiment Analysis - Daniel Mayer, Product Marketing Manager, TEMIS

In this session TEMIS will show how Text Analytics applied to Social Media produces deep Customer & Market Intelligence insights. By applying our solution in real-time to the comments made by online contributors in a key consumer products category, we'll demonstrate live that it has a unique ability to identify, quantify and track consumer preferences for particular brands, products, and features and to illuminate the competitive dynamics at play in the market.

Seamless Social Media & CRM - Bernard Chung, Director of CRM Product Marketing, SAP Labs, LLC

Social media is about people and community, and as social media adoption increases, many companies are making attempts to tap into the social media channels.  As companies learn how best to leverage this growing channel, companies need have a single solution that could not only bring insights into what people and customers are saying, but also be able to seamlessly drive actions based on these insights.  Huge value exists in organization's ability to gather these insights from social media channels and to incorporate them into existing customer management processes within their organizations. Learn how SAP CRM is providing integrated solution like the SAP CRM Twitter solution, that can bring the world of social media and CRM solutions together to work seamlessly; leveraging sentiment analysis to identify who and what your customers are saying, and empowering your service and marketing organizations to take appropriate actions, proactively to resolve issues and to take advantage of opportunities to market through social media channels.

Speakers:
Jeff Catlin, CEO, Lexalytics
Steve Alexander, Senior VP, Serendio
Ravi Condamoor, CEO, Serendio
Justin Langseth, President & CTO, Clarabridge, Inc.
Mattias Tyrberg, CEO & Founder, Saplo
Fiona McNeill, Global Product Marketing Manager - Text Analytics, SAS Inc.
Seth Altman, Senior Consultant, Mark Logic Corporation
Ole-Christoffer Granmo, Associate Professor; Research Fellow, University of Agder; Integrasco
Alec Go, Stanford Univ., Twitter sentiment
Brooke Aker, CEO, Expert System
Sally Church, EVP, Icarus Consultants
Dave Naffis, Cofounder, Intridea, Inc.
Breck Baldwin, President, Alias-i
Ian Hersey, CTO, Attensity
Laurel Earhart, Vice President, Business Development, SentiMetrix
Daniel Mayer, Product Marketing Manager, TEMIS
Bernard Chung, Director of CRM Product Marketing, SAP Labs, LLC, SAP


12:50pm -

Lunch & Affinity Tables


1:50pm -

Selecting a Social Media Analysis Platform/Provider: A Conversation

The market is crowded with specialized tools and services for monitoring and analyzing social media. A few vendors have done a great job of promoting themselves, but lesser-known companies can better meet some requirements. To make an informed decision, buyers need to understand the options and know their own needs. This session will provide an overview of the tools and services in the market, the strategies they support, and the facts buyers need to know before they go shopping.

Moderator: Suresh Vittal, Vice President & Principal Analyst, Forrester Research

Speakers:
Nathan Gilliatt, Principal, Social Target
Marshall Sponder, Social Media Analyst - Web Analyst, webmetricsguru.com AND Porter Novelli


2:35pm -

Pharma Sentiment Surprise! - Online Conversation Yields Unexpected Results

George Bernard Shaw said “The only man who behaved sensibly was my tailor; he took my measurements anew every time he saw me, while all the rest went on with their old measurements and expected them to fit me.” The pharmaceutical industry is now beginning to behave sensibly by actively measuring and analyzing brand and reputation sentiment expressed by consumers online.
This discussion will focus on the framework and implementation of a social media monitoring system to identify an online audience living with epilepsy and analyze their sentiment about treatments to manage their condition. Medication included Vimpat, Keppra, and Zonegran. The analysis of consumer sentiment within online conversations produced actionable information, and established a platform by which the quality of patient care could be improved.

Speaker:
Stephanie Noble, President, Paden Noble Consulting


3:10pm -

Financial Markets

Incorporating News and Sentiment Analysis into Investment and Trading Strategies

News is and always has been a major force in driving financial markets. Moreover, its impact is growing more immediate. Technology is driving the volume and rate of change of the news to the point where it has overwhelmed people's ability to exploit it effectively.


Technology, of course, also has an answer to this problem. It now is possible to incorporate news in the investment and trading process in ways that were not possible just a few years ago. Doing so enables analysts and traders to respond to the ever greater torrent of information faster, more consistently and with increased accuracy. It now is possible to work with breaking news using sentiment and other news analytics which make it possible to better exploit market inefficiencies and more effectively manage event risk.


Machine readable news and sentiment analysis has often been categorized by use only by those most sophisticated firms operating secretive high frequency black box trading strategies.  In this session, we will explore some practical uses applicable across all trading frequencies and short to mid-term investment horizons.

Speaker:
Richard Brown, Global Business Manager, Machine Readable News, Thomson Reuters


3:45pm -

Break


4:00pm -

Voice of the Market

Text analytics can be leveraged in many areas of market research. Tom will give real case study examples of how his firm has merged text analytics with traditional market research and helped fortune 500 clients with customer satisfaction, competitive intelligence, and segmentation.

He will discuss techniques that provide validation through triangulation. Going beyond verbatim concept, themes and negative/positive/neutral sentiment, Anderson Analytics also leverages psychological content analysis which utilize a priori word choice models and compares these to normative, category and demographic specific databases.

Speaker:
Tom Anderson, Founder and Managing Partner, Anderson Analytics


4:35pm -

Sentimental Market Segmentation

The usual goal of sentiment analysis is to provide numeric measures of positive or negative valence for brands, products, and commodities, which can be aggregated over time or geographical regions to analyze patterns and trends. Dr. Shlomo will discuss some new methods he and his team are developing which extend this paradigm in two ways. First, their systems analyze more aspects of each individual sentiment expression, including different types of attitude ("unwieldy" vs. "unreliable"), comparisons ("X is better than Y"), evaluative trends ("X is improving"), and modality ("possibly" vs. "likely" vs. "definitely"). Secondly, they are combining sentiment analysis with their methods for automated authorship profiling, which label texts with author characteristics such as gender, age, native language, education level, and so forth. When this is done, a new type of analysis emerges: data mining can be used to find "sentimental market segments", discovering, for example, that opinion is trending upwards for males aged 20-30, but downwards 30-50 year-olds who did not attend college. He will present some of their research results and discuss the implications for future applications and developments in sentiment analytics.

Speaker:
Shlomo Argamon, Associate Professor of Computer Science, Illinois Institute of Technology


5:10pm -

Sentiment, News, and the Polarity Problem

Description: Although sentiment analysis has a strong history of success on 
customer feedback and certain blogs and editorials, accuracy results are 
mixed for data in the absence of an opinion holder. In particular, news data 
poses some unique challenges to accuracy for sentiment analysis due to the 
blending of what I will call "objective" polarity with opinion-based 
polarity. How is (document-level) Sentiment to be determined, for example, 
in an article about the Haitian earthquake that discusses humanitarian aid? 
Similarly, an article about Bernard Madoff’s jail sentence shows a highly 
negative “objective polarity” somehow mitigated by a subsequent action. And 
how can we tease an author’s opinion from the semantics of objective polarity 
where they exist in news data? Author opinion (often referred to as “bias”) 
in news data is subtle in its indication by design. This talk will also 
discuss the grounding of the concept of "sentiment" within the greater 
context of the Semantics of Opposition.

Speaker:
Leslie Barrett, Consultant, LBTechconsulting


5:45pm -

Conclusion

Speaker:
Seth Grimes, Alta Plana