Carnegie mellon sentiment analyses overview
1 like710 views
The document provides an overview of sentiment analysis and its applications, highlighting key practices in digital marketing and social engagement led by experts like Susan Emerick and Amy A. Laine at IBM. It emphasizes the importance of keyword definition and data analysis in extracting insights from social media and online content to inform marketing strategies. The document also discusses trends in brand reputation management and the integration of social media dialogue into customer relationship management systems.
Carnegie mellon sentiment analyses overview
- 1. Sentiment Analysis Overview October 12, 2012 ยฉ 2010 IBM Corporation
- 2. Susan Emerick Program Manager, Social Business Enablement Digital & Social Influence Strategy & Development, IBM CHQ Susan Emerick is a seasoned integrated marketing communications consultant with deep expertise in Digital & Social Influence Marketing Strategy. She has a proven track record developing effective marketing programs utilizing the best mix of on-line marketing techniques, exploiting new market opportunities created by the worldwide adoption of social media, mobile and other emerging technologies. In her current role, Susan is dedicated to evolving marketing as a practice, articulating the benefits of integrating digital and social influence programs to foster long-term, high value relationships with clients, prospects, partners, colleagues, and communities. She currently leads many of IBMโs transformational workforce initiatives which empower IBMers across the globe to deliver business value through sharing their expertise across the social web. Beginning in 2008, Susan was instrumental in establishing IBMโs C.O.R.E. (Cross functional, On-going Research & Engagement) Social Marketing practice in partnership with the IBM Market Insights team. This social marketing methodology was founded on gathering social โlisteningโ intelligence through on-line research and applying key insights to marketing planning and social engagement strategies. This IBM initiative was awarded the 2010 SAMMY (Social Advertising, Media and Marketing) for โBest Socialized Businessโ Susan enjoys sharing her expertise with other marketing professionals by speaking at a range of leading industry conferences including, Word of Mouth Marketing Association (WOMMA) School of WOM, iMedia Summit and DMAD. Prior to joining IBM, Susan led the development of global marketing programs for both B2B & Consumer brands across various industries including Financial Services, Media & Entertainment and Retail Distribution. LinkedIn: www.linkedin.com/in/sfemerick Twitter: @sfemerick Blog: www.susanemerick.com 2 ยฉ 2010 IBM Corporation
- 3. Amy A. Laine Principal Market Analyst, Team Lead Client Research, IBM CHQ Market Insights Amy A Laine is a Principal Market Analyst and Team Lead within IBM's Market Insights division. She leads the Market Trends and New Opportunities Program, and is currently driving targeted research to best inform business decisions as IBM embraces the digital marketplace. Amy is a founding partner in the design of IBMโs C.O.R.E. (Cross functional, On-going Research & Engagement) social media program โ a program that embeds research as the foundation for an enterprise-wide platform for a social media strategy built on planning and engagement. Having been with IBM for over a decade, always within the market intelligence/market insights division, Amy believes that starting with research and incorporating continued measurement throughout the execution or active outreach phases is critical for a successful social strategy meant to transform the business. 3 ยฉ 2010 IBM Corporation
- 4. Sentiment Analysis: What is it and General Uses What is sentiment analysis? โAutomated sentiment analysis is the process of training a computer to identify sentiment within content through Natural Language Processing (NLP). Various sentiment measurement platforms employ different techniques and statistical methodologies to evaluate sentiment across the web. Some rely 100% on automated sentiment, some employ humans to analyze sentiment, and some use a hybrid system.โ - Maria Ogneva is the Director of Social Media at General uses for sentiment analysis โ Brand Health Monitoring โ Competitive Positioning โ Predictive Analytics and Modeling โ Customer Advocacy: โข Customer value โข Customer attribution โ Customer Pain Points โ Marketing planning and segmentation โ Campaign effectiveness Biz360 4 ยฉ 2010 IBM Corporation
- 5. There are several key steps to harvesting insights from the Digital Marketplaceโฆ DEFINE keywords MINE publicly available social media data within specific dates based on keyword relevance FORUMS / NEWSGROUPS VIDEO SHARING ANALYZE data via human analysts WIKIS ESTABLISH benchmark metrics MICROBLOGS Sentiment analysis of publicly available content SOCIAL NETWORKS SOCIAL MEDIA NEWS AGGREGATORS PHOTO SHARING * OR * EVALUATE performance against benchmark BLOGS 5 ยฉ 2010 IBM Corporation
- 6. The first step, keyword definition, is of critical importance to the quality of the insights - and decision-making - based on the data In a world with a billion computers, four billion cell phones and a robust global Internet, there is an overwhelming amount of digital messages being posted online every day Most are not relevant to your brand or specific topic of interest. Within those that are, not all are relevant Establishing pre-defined keywords allow us to narrow down the universe of all possible posts to only those that are relevant to our research needs Much like developing a screener to determine who you want to invite to a focus group (e.g., โLarge EnterpriseโโIT professionalsโ who are โhardware purchase decision-makersโ in the โU.S.โ), we need to determine the criteria for inclusion in the listening sample set by defining the keywords that signal: Include this POST in data collection If the keywords are too broad, then we get โnoiseโ (i.e., irrelevant posts) If they are too narrow, then we miss relevant conversation and may draw erroneous conclusions Filter Universe of all posts EVERY POST 6 Include Unique Posts RELEVANT POSTS ยฉ 2010 IBM Corporation
- 7. Keywords often need to be refined and qualified in different waysโฆ For conversation mining, several โstringsโ can be employed: โ A category string designed to pull in discussion relevant to a specific server โ A branded string designed to pull in mentions of IBM within the larger server discussion โ A category string designed to pull in mentions of specific products within the larger server discussion The category string is shaped into a Boolean keyword string* By zeroing in on the terminology that buyers and decision makers actually use, we will best capture their online conversations Server A IBM + Server A Product T Product U Server B Server B + Product V Produce W Product X Product Y Server C Server C *Note: Boolean Keyword String โ A set of keywords that employs Boolean logic to focus and return specific , relevant messages in search 7 ยฉ 2010 IBM Corporation
- 8. The analysis focuses on what you want to know. The overall conversational volume will show peaks during product launches Volume 1Q 2Q 3Q โข Volume is based on keyword matches โข Volume is measured on the record level 8 ยฉ 2010 IBM Corporation
- 9. But who is contributing to the conversation is of critical importance when analyzing the online discussion as well โVoiceโ Media / Press Professionals 1Q 2Q 3Q Employees Consumers Investors Competitors Analysts Business Partners Executives 9 ยฉ 2010 IBM Corporation
- 10. And it is not only how much is said, but why people are contributing? Conversation Volume by Message Type 10 10 IBM Confidential ยฉ 2009 IBM Corporation
- 11. And where they are talkingโฆ Conversation Volume by Venue 11 IBM Confidential ยฉ 2010 IBM Corporation
- 12. It is also valuable to understand how conversational themes are related to one another. Conversation Topic Relationships, Volume and Sentiment Server B Performance User Preference Product V Performance Events Product W Interoperability Industry News Announcements Cost/Affordability Product T Support Server B 12 12 Source: Converseon, 2010 Sentiment Analysis Research Study IBM Confidential Security Server C Server Migration Scalability ยฉ 2009 IBM Corporation
- 13. Word clouds provide an understanding of the most often mentioned terms within the relevant buzz Word Cloud of General Online Conversation โข Word size corresponds with frequency of occurrence within the data set 13 Source: Converseon, 2010 Sentiment Analysis Research Study ยฉ 2009 IBM Corporation
- 14. Closing thoughts on top trends and whatโs on the horizon โฆ. Influence: out of all the posts about your brand, how do you pick the top 50 to focus on and connect with - relationship management and priority coverage modeling Reputation management: Establishing prominence, reputation management - focused enablement of employees to build influence as a reputable authority in relevant conversation Enterprise wide response management: You may hear this referred to as Social CRM, tying social dialogue into customer relationship management systems Real time, Predictive, Actionable, anticipating customers needs and be more responsive to improved customer experience Business Transformation, from social media primarily considered โconsumer oriented networkingโ to applications in Social Collaboration and Social Networking to achieve Business outcomes 14 IBM Confidential ยฉ 2010 IBM Corporation
- 15. Thank You ! 15 IBM Confidential ยฉ 2010 IBM Corporation
- 16. Open Forum : Q&A 16 IBM Confidential ยฉ 2010 IBM Corporation