A Measure of Emotion Dear Watson
Sentiment Analysis and IBM Watson
“I can calculate the motion of heavenly bodies but not the madness of people.” Isaac Newton
People can distinguish criminals from non-criminals just by looking at photos. Machine learning neural networks have been shown to make accurate inferences based on the curvature of the upper lip, the distance between the two inner corners of the eye and the angle drawn from the tip of the nose to the corners of the mouth. It is also possible for machines to be trained to look at faces and make the same conclusions as people, biased or not, on values of trustworthiness and dominance. Whilst these results raise some serious ethical considerations they are interesting applications of artificial intelligence.
My team has no current need to categorise using images. We do however see value in categorising and understanding the emotional tone of written messages that we receive on various platforms. This can allow us to adopt our customer engagement approach and identify areas of improvement.
Sentiment Scores & Magnitude
Sentiment analysis, also known as Emotion AI or Opinion Analysis, is the identification and categorisation of emotion in words through the use of text analysis and computational linguistics.
At a basic level the sentiment score of a piece of text or document indicates the positive or negative emotion of the text. Some API’s like IBM Watson Tone Analyser provide a more detailed breakdown of the various types of emotion such as anger, fear, satisfaction. The magnitude score indicates how much emotional content is present and can be proportional to the amount of text analysed.
Sentiment analysis can be used to analyse emotions and tones in email, reviews, tweets, Facebook wall posts, messages or company reviews. You can also enhance customer service interactions through responding according to emotional intensity. There are also possibilities to integrate with chatbots and to build dialog strategies that adjust according to the emotion measured.
The Platform — Watson Tone Analyser
We have used Watson Tone Analyser in our proof of concept build for customer service inbound messages.
We can measure the emotional content of requests that are sent to one of our support teams. This will have a direct impact on the calculated priority, specifically the urgency rating assigned to those requests. If a threshold is crossed for the anger, fear or frustration emotions then the urgency rating of that ticket is increased. This provides an additional criteria to filter our service queue and one that adds value to the more standard measures of segment, severity, days open.
The Solution
We have integrated Tone Analyser into Salesforce & Facebook Messenger with Stamplay as the automation platform.
In this video you can see an overview of sentiment analysis in action created by our very talented senior application developer Gnani Matavalam.
A potential next step is automatic sentiment analysis and speech to text transcription from video. How about each Suitebox meeting being automatically transcribed with topic breakdown and summary for financial advisers?
What Next?
Extending on sentiment analysis, personality insights provide details on how and why people act, think and feel the way they do. This is an area we are currently exploring via IBM Personality Insights which applies personality theory and linguistic analysis to infer attributes from text. Understanding potential clients values, needs and preferences could inform and improve many interactions.
Human judgement is a key skill that will not be replaced by machines. However AI solutions can offer us shortcuts to inform that judgement, such as emotional tone and trends related to client interactions and online opinions.
The purpose here is to share simple practical solutions on accessible platforms that could be useful to test out your ideas. It’s not to delve into lots of technical detail and compare options. If you want to see what else you can achieve via IBM Watson check out this link. If you would like to explore Tone Analyser in more detail, this link is a good place to start.
Thanks to Lucy Adelaide for working on the illustrations above and the talented Gnani Matavalam for building the demo.
The views expressed in this article are those of mine alone. This is part 3 of a 6 part series exploring practical business applications of artificial intelligence.
Originally published at https://www.linkedin.com on October 25, 2017.
