Using Sentiment Analysis For Social Media

What is Sentiment Analysis in Social Media?

The technique known as sentiment analysis is a way to extract subjective sentiment information from a source of data. Sentiment refers to how a person feels towards a product or topic, and can range from positive to negative.

This can be an extremely useful tool to gauge users’ attitudes and feelings towards a brand, product or event, based on their posts to social media platforms (e.g. Twitter, Facebook etc.). We’ll take a look at some important considerations when using sentiment analysis in order to make the method as valuable and insightful as possible for a design research project.

‍

‍

Is it the Right Method?

Before diving into this type of method, ensure it is adequately suited to your project and what you are trying to discover. While it can be an extremely useful method to understand sentiment based on social media postings, this method might not be appropriate in some instances.

Looking at the product or topic you are looking to get insight on, does it have a large and active user base on social media platforms? Will social media accurately represent the audience, or will it only represent a small portion of users that will engage with social media?

Social Sentiment Analysis Techniques

Automated vs. Manual Sentiment Analysis

Some programs exist that offer manual sentiment analysis, using Natural Language Processing techniques, Salesforce Radian6 is one such program.

Pros:

• Automating the process means data analysis is faster
• Can analyse extremely large data sets
• Can generate a visual dashboard of results which is easier on the eyes than looking at lots of data

Cons:

• Automated process not as reliable as human coding (e.g. not as able to identify instances where users are being sarcastic)
• Programs can be expensive
• Need to have access to the social media account e.g. Twitter login details which some companies are not keen to share.
• The visual dashboard can look quite clunky

‍

‍

Sentiment Analysis Coding

Although manual sentiment is thought to be more reliable than automated, it also involves considerable more time to go through the data set and manually code it. Think of ways to combine both automated and manual techniques to find the best method to extract meaningful insight.

‍

‍

What to code for

Besides purely looking at the overall sentiment in a statement (e.g. positive, negative, neutral), you should also be coding the qualifying subject of that sentiment (e.g. what is the source of that sentiment). It can also be useful to further break down general sentiment into more specific emotions to get a better understanding of how users are reacting to a specific topic.

The more you can take into account when coding, the richer your analysis will be in offering a complete view of emotional attitudes towards a product or topic. Depending on the subject, it may be relevant to look at competitor product mentions, for instance.