In-Scope: Why sentiment analysis needs to explore more than just positive, neutral and negative.

Briony Lewis is a Data Analyst for OMD Create Melbourne, helping to unlock data leading to robust insights and more effective campaigns. 

Sentiment analysis is a powerful tool used in many companies to interpret customer attitudes and behaviours. Using sentiment analysis can help you navigate the reason behind affinities to one brand versus another or determine what categories resonate most with people and why. Whilst there are a lot of strong analysis tools out there, it is important to consider customisation and machine learning capabilities that allow you to apply different linguistic rules and go beyond the basics.

Sentiment analysis technology often uses machine automated algorithms, which take in various contexts, sentiment dictionaries and linguistic rules to identify comments as neutral, positive or negative. Whilst this is important for looking at the big picture, there are many drivers behind sentiment thus analysis becomes more impactful when sentiment is categorised into emotions such as joy or anger, opinions such as product feedback or reviews, and subjects such as food or travel categories.

Say for example you’re a travel client. Negative sentiment can have multiple layers to it. One customer could be complaining about the service they received, whilst another could be angry at the load time and usability of the web page. The former comment is relevant to a customer service representative, whilst the latter could in fact drive the SEO team to make improvements to these landing pages. In both cases, sentiment is subjective and relevant to different teams meaning there is a need to be customisable.

This is where machine learning comes in. Certain tools will offer features that apply pattern-based learning. This is more sophisticated than the usual algorithms as it allows you to categorise comments into emotions or personalised topics such as ‘Negative: UX experience’ & ‘Negative: product review’. This is an important feature to consider when you’re looking at a tool because human language is nuanced and complex. Context is hard to interpret and a human’s perspective of positive will most likely differ from that of a machine taught to just weight positive words against negative and neutral.

So, next time you are deciding on which tools to have in your company ask these following questions:

1. Are there capabilities to search and filter through sentiment?

2. Can I manually change sentiment?

3. Can the tool be trained to customise sentiment categories?

4. What is the average sentiment accuracy with your software?

When it comes to sentiment analysis, tools and technology are paramount to success in marketing. Here at OMD, we understand that the added human input and machine learning is what uncovers more robust insights, setting you apart from your competitors as original thinkers.

Loading Facebook Comments ...

Leave a Reply

Your email address will not be published. Required fields are marked *