Sentiment Analysis -The Next Big Thing In Social Media Marketing
Posted in Marketing & Strategy Articles, Total Reads: 3671
, Published on 31 October 2013
Sentiment Analysis is defined as the task of finding the opinions of authors about specific entities. The decision making capabilities of people are affected by the opinion and thought of people around the globe. The best place to capture the global opinion of the people is the social media. It contains plethora of data pertaining to different thoughts that people have while using a particular product. There is an explosion today of the ‘sentiments’ available from social media including Twitter, Facebook, message boards and user forums. The snippets of texts present are a gold mine for companies and individuals that want to monitor their reputation and get timely feedback about their products and actions. When customers interact with the company through their products or services, they put a voice of theirs on the social media channels like Facebook, Twitter, Foursquare etc. Sentiment analysis helps to capture customer views on this sort and do a client research to calculate whether an opinion expressed is mainly positive or negative. To the marketer, this data provides deep insights into consumer behavioral tendencies and present an opportunity to learn about customer feelings and perceptions in real time, as they occur without intrusion or provocation.
Types Of Sentiment Analysis: It can be classified into
Sentence Level Sentiment Analysis: This is the simplest form of sentiment analysis and it is assumed that the sentence contains an opinion on one main object expressed by the author.
Document Level Sentiment Analysis: In this format, multiple sentences are clubbed together to form a paragraph and finally a document. The scoring for the document is done by segregating documents into paragraphs and paragraphs into sentences. A sentence level score is calculated which is cumulated to arrive at the final score for the paragraph. Some sentences might talk positive about the product while some may negate the positives in concluding sentences. Hence a cumulative score is the sum of all the scores to arrive at a single score for the paragraph and hence the document.
Aspect Level Sentiment Analysis: in many cases people talk about entities that have many aspects (attributes) and they have a different opinion about each of the aspects. This often happens in reviews about products or in discussion forums dedicated to specific product categories (such as cars, cameras, smartphones, and even pharmaceutical drugs).
METHODOLOGY: To measure sentiment on a larger scale, brands typically use software that applies Natural Language Processing (NLP). Although the exact methods used by each software may differ slightly, the ultimate goal of NLP technology is to process and understand human language and automatically derive meaning. Polarity of a given text is either positive, negative or neutral. Classifying text according to polarity is a difficult task. The polarity is represented in a user readable format. It may be in the form of graphs, drawings, etc.
Thus through complex open source resources like Natural Language Toolkit , Stanford NLP Group we can create a framework for calculating sentiments of sentences to arrive at a conclusion for the data in hand.
USAGE IN INDUSTRY: Various tools available in the market are Radian6 (salesforce.com), Cymfony (Visible Technologies), ScoutLabs (Lithium Technologies). Some of the bigger known IT giants like TCS, Infosys, IBM also have their customized tools for doing a client based research and brand wise analysis and have generated huge amounts of revenue through these platforms.
Companies like Starbucks and American Cancer Society (ACS) have been using sentiment analysis for reading your feeling and opinions online about itself. Did you know that there are 10 tweets every second mentioning Starbucks? The American Cancer Society gets about 6,000 mentions a month on Twitter, public Facebook pages, blog posts, and in comments on blogs or articles. FMCG giants Unilever had employed the solution CrowdFlower had to offer. The percentage of content in which positive or negative sentiment was identified rose from 15% to >95%. The accuracy rate rose from less than 30 % to more than 90 %.
Best Buy (BBY), Viacom's (VIA.B) Paramount Pictures, Cisco Systems (CSCO), and Intuit (INTU) are also using sentiment analysis to determine how customers, employees, and investors are feeling. In one of the case studies, Verizon which was considered a fee happy company with consumers, due to a change in its policies regarding fee charges, faced the consumer’s wrath as they didn’t find it as welcome news. Crimson Hexagon analyzed over 4,000 reactions on Twitter, and while the reaction was (as expected) resoundingly negative at 51%, it’s interesting to identify what specifically those on Twitter were griping about (what’s driving the sentiment).
Brand Monitoring: Evaluate and monitor your brand on the basis of customer views and opinions.
Campaign Monitoring: Create product campaigns to gain insight about the perception of the people on social media about the product since it spreads virally.
Competitive Intelligence: Track and follow your competitors and their customers to check their pain points and areas of improvements to tap their potential.
Identifiers: Check the influencers who are talking about your brand in various channels and influencing others.
CHALLENGES FACED: There are various challenges faced while calculating the sentiment of a sentiment. Some of them are illustrated below:
Context of the sentence: A major challenge faced while capturing the essence of the data is when there is a difference between the actual context and the implied context of the sentence. It is very difficult to find the context when the sentence could be positive in nature but the implied context could be negative. Example: “Sarcasm is so cool these days”.
Use of Slangs and other Languages: Social media data is full of Hinglish language and slangs. It is sometimes difficult to deal with such kind of data since the dictionary which is being used to calculation might not contain these types of words rendering them neutral.
Sentiment for multiple Attributes: When users’ posts their reviews about a product like a phone suppose an Iphone, they might speak positively about certain features like screen, processor, interface while negate several things like battery, touch screen and screen size. Now the opinions are actually attached to specific features of the iphone. In these cases it is difficult to determine the sentiment as a whole for the entire text available.
Sentiment Analysis has been helpful in giving us an accuracy of 60 to 70% which is actually comparable with a human analyst. It can become a fuel for efforts to shape opinions, attitude and emotions on the web and hence can be used for predictive modeling. It can be used for marketing strategies, Poll Prediction, Government Intelligence, Product Comparison and other numerous business activities to come out with a result which has a bearing on the decision making body. In stark contrast to marketing of the past, today’s marketers are measured by how much revenue they bring in per dollar spent. What “sentiment analysis” does is give those marketers an alternative way to measure their effectiveness — tracking how customers feel about and how much they are talking about a brand. All that effort on branding, messages, and color schemes can finally be validated! However, the number of positive mentions you generate is simply a step towards revenue. If more people are talking about Pepsi than Coke, but more people are buying Coke than Pepsi, there is something seriously wrong in the conversion funnel towards a purchase. Thus marketers can track their campaign’s performance in real time and get can idea about the general consensus in the market and from hindsight can predict how their future decisions would lead to the conversion of consumer’s money into our own bank balance.
This article has been authored by Saurabh Jain from Goa Institute Of Management
1. Sentiment Analysis By Ronen Feldmen
2. Sentiment Analysis: A Perspective on its Past, Present and Future-I.J. Intelligent Systems and Applications, 2012, 10, 1-14
3. CrowdFlower Case Study - http://crowdflower.com/docs/CF-Unilever-CS.pdf