There is No Why in Sentiment Analysis
‘Why’ – the number one three letter word that every parent gets to hear a million times when their child reaches the age of unlimited curiosity and looks to their parents to deliver the voice of knowledge behind each and everything. And even when mom can give a perfect answer to an impossible question, she will quickly find herself questioned with a follow up ‘why’.
Marketers need to ask themselves the same question for their consumers. Why are people using the product the way they are? Why aren’t they buying my brand? It is not enough to just understand what is happening in the market place. More importantly marketers need to understand the ‘why’ behind consumer behaviors. It is only by answering that question that a marketer will ultimately identify a core consumer insight that builds the basis of every successful marketing campaign.
Up until shortly, companies needed to spend significant amount of resources to collect a small set of consumers to conduct focus groups and surveys in pursuit of consumer insights. With the rise of the digital age and social media, things changed. Companies now have the ability to gather more data than ever before on a large number of brand users in near real time and at a much lower cost with the help of Social Media Monitoring Systems (SMMS).
An important tool that is standard in many SMMS is sentiment analysis. It allows the user to determine the sentiment of thousands or even millions of posts by classifying them as positive, negative or neutral. The analysis has proven to be vital for marketers and CRM teams alike in dissecting the online activity or in gaging a brand’s social reputation or social campaign success. But despite its undisputable benefits, the classification of data by sentiment represents only the first step towards discovering the consumer insight that marketers all seek to find.
Merriam-Webster defines sentiment as “an attitude, thought, or judgment prompted by feeling”. Therefore in our pursuit of ‘why’, we should take a closer look at the feelings or emotions in play that define the sentiment. For example, though two consumer reviews might be classified as positive, the type of emotions being used to express that sentiment can vary. Someone posting “I like your product” is emotionally dissimilar to someone posting “I have never before experienced such a great product”. Both reviews would be categorize as positive in a sentiment analysis, but if actually identified separately would have a different importance for the respective marketer or CRM team.
Here is another example. Both “I am not sure about this brand” and “I hate this brand” would be categorized as a negative sentiment, however as most people might agree, hate is a much stronger emotional expression vs. uncertainty and requires a different approach in addressing the consumers attitude. This is where emotion analysis comes into play.
Soshio’s emotion analysis examines the individual emotions expressed in each post by applying a point system – lets call it passion points for simplicity reasons – providing the user with an indication of which key emotion is being expressed surrounding a specific topic.
At Soshio we categorize each emotion based on the ‘Big Six’ defined by Paul Ektman and added ‘Desire’ to the list to make it the ‘Soshio Seven’:
- Joy: General mood of happiness, laughter, and fun.
- Love: Strongly positive reactions toward an entity or experience.
- Desire: Strong longing for an entity or experience.
- Anger: General mood of rage, cursing, and angst.
- Fear: General mood of horror, paranoia, and worries.
- Loath: Strongly negative reaction toward an entity or experience.
- Sorrow: General mood of sadness, weeping, and melancholy.
With the ability to identify explicit emotions, companies can track and engage with consumers expressing a very specific feeling. So instead of having a CRM team monitor all posts labeled as negative, they are able to prioritize by the importance of each emotion, saving them time in the process.
Similar applies for marketers. Marketing teams can track and see if their campaign is triggering the desired emotion instead of simply gaging a campaign’s success by the number of positive sentiments.
Having said that, Emotion Analysis does have its challenges.
Identifying feelings and emotions through a sequence of words can be difficult, especially when taking the cultural context into account. Irony, sarcasm etc. are hard to identify and requires sophisticated algorithms that are still being tested.
Also, the emotional interpretation of a post can vary based on the cultural heritage or even emotional state of the user. A data point could be tagged as ‘desire’ by the tool, however interpreted as ‘love’ by the user, thus raising the discussion of data accuracy.
Despite the challenges, the opportunities that Emotion Analysis provides are far from few and with companies needing to hone in as precise as possible on their target audience, the need to move beyond a polarity-based analysis of positive and negative is certain.
More Business articles from Business 2 Community: