Sentiment analysis (sometimes known as opinion mining or emotion AI) is when software analyzes text and determines whether it expresses positive or negative sentiment. It’s often used on customer feedback like reviews and survey responses, tweets, and interview transcripts.
Products that offer sentiment analysis usually analyze the sentiment of the whole document. Since this is a roll-up, it can be very inaccurate.
In Dovetail, we analyze the sentiment of sentences, rather than a roll-up of the whole document. This allows for much more granularity, and highlights core pain points or compliments hidden in the document.
As an example, an online flight review that covers several topics can be broken down and tagged appropriately:
“We were flying to/from LHR via YUL. These seats are located immediately behind a service bulkhead and afford a lot of legroom. The remote control is on wired tether which often tangled with other items and accidentally turned lights on or off or called cabin crew unnecessarily — annoying. The seats are leather-faced and comfortable. An amenities kit is provided with pillow and blanket.”
Dovetail will break this review into sentences and tag each appropriately. The bit about the remote control will be tagged negative, while compliments about the leather-faced seats and amenities will be tagged positive.
Sentiment analysis becomes quite useful when dealing with large amounts of unstructured data like customer reviews and user feedback. Not only does it help you quickly gain an understanding of the opinions expressed across a bunch of data, but it also gives you a great starting point for further tagging.
Like everything in Dovetail, sentiment analysis is easy to use. In under two minutes, researchers and product managers can sign up, confirm their account, and analyze sentiment on their data. From there, it’s straightforward to use our tagging system to apply more granular tags.
Here’s a quick demo:
Automatic sentiment analysis is available now to all customers. It can be enabled when you import data using any of the following methods: