DIY Social Sentiment Analysis in 3 Steps Feb 26, 2015
About NetElixir Mission: To Help Digital Marketers Succeed Online. Incorporated: 2005. Global Offices: Princeton (HQ). London. Hyderabad. Team: 75+ fanatically analytical search marketers with over 5.5 MM hours of hands-on retail search marketing experience.
Client List Within 12 months of implementing our strategy, organic traffic more than doubled & transactions went up significantly. Ken Bausch, VP Interactive Marketing, World Kitchen 5
About NetElixir University NetElixir University was launched in 2012 with a vision of democratizing the digital marketing industry through exceptional knowledge and expertise sharing. Our goal is to share the best practices in retail + digital marketing with 10,000 businesses worldwide by 2015. NetElixir University 6
About Udayan Bose Founder & CEO, NetElixir. Founded PartyBingo.com (PartyGaming, Plc). Guest Lecturer: Johnson School of Management, Cornell University. City University of New York, Baruch. Indian School of Business. http://www.linkedin.com/in/udayanbose udayan@netelixir.com 7
TOPICS INTRODUCTION TO SENTIMENT ANALYSIS VALENTINE S DAY TWEETALYTICS: NETELIXIR RESEARCH OBSERVATIONS
9
What is Sentiment Analysis Sentiment analysis aims to determine the attitude of a speaker or a writer with respect to some topic or the overall contextual polarity of a document. A basic task in sentiment analysis is classifying the polarity of a given text at the document, sentence, or feature/aspect level whether the expressed opinion in a document, a sentence or an entity feature/aspect is positive, negative, or neutral. Advanced, "beyond polarity" sentiment classification looks, for instance, at emotional states such as "angry," "sad," and "happy." 10
Social Media: Sentiment Analytics How we get a pulse on what is happening around us Using sophisticated tools, we captured and analyzed structured & unstructured data from social media to extract valuable insights We decided to focus on Twitter because of the way in which emotions are captured in the140 character limit
Social Sentiment Analysis Social Analytics even captures the mood of the nation! Researchers from University of Bristol collected 484 million tweets from 9.8 million Twitter users. Then they extracted key words from tweets and algorithmically assigned a mood to each word. Then they calculated each word s frequency and determined U.K. s overall mood from the frequency of different words on different days.
Social Sentiment Analysis Social Analytics even captures the mood of the nation! 2011 Summer Twitter Mood Anger Royal Wedding Twitter Mood Happiness
14
How to derive insights from social sentiment analysis How can we capture relevant data and extract useful insights? Insights Text Analytics = Data Pre-Processing + Data Analytics + Semantics + Scoring
Valentine s Day Tweetalytics A NetElixir Original Study on Valentine s Day tweets to uncover sentiments & trends Data Collection Collect Valentine s Day Tweets Data Processing Tweets cleansed and processed through R Data Analysis Data Analysis done to find out trends/sentiments Note: Mean Absolute Percentage Error of the Model = 6.95% (Model is able to accurately classify at 93.05%)
Step 1 Data Collection: Collect tweets with relevant hashtags Tweets with Valentine s Day hashtags were selected (e.g.:#valentineday, #valentinesday) Hashtag Selection A script was written to capture the tweets with the selected hashtags from U.S geo on a daily basis Data Capturing The collected tweets were stored in a specific internal server instance Data storage Tweets were collected from Feb 9 th till Feb 15 th Data capturing rate was @ 425 Tweets / Hour
Step 1a Data Pre-Processing: Filter out irrelevant tweets 65% of the tweets belonged to marketers Based on retweet count, presence of promo text and other patterns, marketer s tweets were identified. Majority of the marketer s tweets were removed to attain a resulting data set which predominantly consisted of regular user s tweets 25,000+ Tweets
What Are you Doing to Drive Engagement? 62% of Marketer Tweets = 0 engagement Is becoming more of a sales channel than an engagement platform?
Step 2 Data Processing: Import the data into a statistical computing tool The collected tweets would be imported into R 25,000+ Tweets Once we imported the data into R, data cleansing was performed using pre-defined libraries Data Stemming converting words to their root forms (e.g.: loved love) Convert to lowercase, remove punctuations, usernames & links R, Free Software for Statistical Computing Tweet data set, ready for analysis Remove stop words (e.g.: the )
Step 3 Data Analysis: Apply sentiment analysis & word frequency algorithm Applied sentiment analysis and word-frequency algorithms pre-defined in R upon the cleansed data set
22
Top Valentines Day Activities V-Day dinner Couples want to listen to Spotify together Watch Galentine s Day or 50 Shades of Grey Starbucks Coffee Tour to Disneyland Manicure & Nail-art before V-Day
Most Talked About Valentines Day Products
Chocolates: Extreme Sentiment Polarity Though most of the chocolaty tweets are in the neutral segment, Advanced sentiment analysis showed that 77.3% of the positive chocolate tweets were Extremely Positive I not only have the best valentine, but I now have the best chocolate #valentinesday No #ValentinesDay this year. Oh! I miss those yummy rummy chocolate
Concluding Remarks Social Sentiment Analysis is a great gauge of how connected customers are with your brand. Advanced Sentiment Polarity Analysis helps you get a better sense of what your most passionate advocates and top critics are saying about your brands. Though, to uncover a trend / pattern, Twitter alone may not be sufficient.
Request a step-by-step Technical DIY Sentiment Analysis Process Chart { marinn@netelixir.com Follow Us @NetElixir 27
{