I am working on a paper and a presentation on sentiment analysis, and decided to put together this very short overview for you (procrastination, anyone?)
Sentiment analysis is one of those topics permeating every area of a marketers’ life. The other day, a colleague even mentioned that he was doing some sentiment analysis of social media conversations for a court case on competition!
I hope this short overview will help you understand what on earth sentiment analysis is, and decide whether you need to learn more about how it’s done so that you can use it in your own work.
Why study sentiment?
How and what we feel impacts every aspect of consumption: from information retrieval, processing and retention, to decision-making, behaviour and even the assessment of consumption experiences. I think we can all relate to being less patient and tolerant of delays or mistakes on, say, a Monday morning than, say, on a lazy Saturday afternoon 😉
Image: The Guardian
What are we talking about when we talk about sentiment analysis?
Sentiment analysis consists of a number of techniques – and, increasingly, technology – to identify and categorise feelings.
Typically, we want to find out:
- Whether consumers expressed positive or negative emotions
- How strong that sentiment is
Here is an example relating to the airline industry in India:
In some cases, we also want to find out the exact type of sentiment, as that will influence how people behave and/or what they will value the most. For instance, does the text reveal a broad, passive sentiment like “sadness”, or a targeted, active one like “anger”?
How can we identify and analyse emotions?
Traditionally, this has been done via experiments – think about psychology’s many mood induction and manipulation studies, mostly featuring undergraduate university students. It can also be done with interviews or surveys about previous emotionally charged events. Recently, social media emerged as a promising source of input for sentiment analysis, as we share so much information in these platforms about ourselves, what we do and what we think.
If using social media data, the first step is to collect the data (e.g., tweets or blog posts) from the relevant platforms using content scrapping software, into which you enter your search parameters, such as selected keywords.
Next, we typically look for sentiment polarity – i.e., whether the overall feeling is positive or negative (or, indeed, neutral). This is done by looking for particular expressions that reflect the state of mind of the person that wrote or said something.
In some cases, we also want to identify the specific emotion experienced. Again, this is done by scanning the text and picking up expressions or phrases that denote a particular sentiment.
The techniques for picking up those expressions are based on semantic analysis and language processing. It is a fascinating field, and a rapidly evolving one (e.g. relating to imagery) of which, unfortunately, I (still) know very little.
What tips and resources for beginners would you add to this overview?