In the involve.ai platform, Sentiment refers to the mood or emotions being expressed in your qualitative customer data. Our AI models recognize words as positive, negative, or neutral and assign overall sentiment to every qualitative interaction.
Sentiment is an extremely important and often overlooked piece of the data puzzle. Quantitative data tells a compelling story, but qualitative data, and the emotion behind it, provides essential context. Identifying interactions with intense sentiment early helps organizations address concerns and optimize successes before those feelings trickle into the quantitative data.
How it works
Our models take unstructured text (for example: emails, meeting transcripts, notes, webinars, support tickets) and assign a polarity score to each and every word in the text based on their orientation and intensity. This score is a value between -1 and 1, where -1 is extremely negative, 1 is extremely positive, and 0 is neutral.
After calculating the individual scores of all words in the text, a pooling operation is applied to assign polarity to the entire interaction, along with all other interactions being calculated.
involve.ai then groups each of these interactions per customer and calculates the average polarity. Average polarity per company is normalized to High (2), Medium (1), or Low (0) and is shown on the Dashboard under Customer Sentiment.
The normalized values are one of seven inputs that comprise the involve.ai Customer Health Score.
Reviewing sentiment inputs
If you want to better understand a specific customer’s sentiment score, scroll to the right, to that customers’ involve.ai Customer Health Score and click the > to open additional details. In this side panel, select History, and you can scroll through to see the sentiment assigned to individual interactions.