How many of us have gotten irritated, frustrated, or angry during an automated customer service call? While a human representative answering the phone might be able to pick up on a customer’s tone, the automated messages, listing options for which department we want to reach, or what exactly our issue is, often seem to disregard our feelings completely.
This may soon change. The Tone Analyzer for Customer Engagement from IBM Watson attempts to bridge the gap between the human and automated sides of the conversation by expanding how it interprets tone of voice. According to a blog post on Thursday by Rama Akkiraju, distinguished engineer and master inventor at IBM Watson, the tool can pick up on seven different types of tone via conversations with customer service agents and chatbots: frustration, satisfaction, excitement, politeness, impoliteness, sadness and sympathy.
And beyond that, the Tone Analyzer also claims to be able to detect these sentiments in emojis, emoticons, and slang.
The Tone Analyzer was developed by using a machine learning algorithm that trained on customer support conversations on Twitter. It also detects tone throughout the conversation, noting how it changes and giving feedback to agents suggesting when they should become more sympathetic, polite, or excited to help smooth out the interaction.
“The new feature makes a chatbot tone-aware, enabling it to provide unique responses to frustrated, sad, or satisfied customers,’ the blog post stated. “For example, it can respond to sadness with, ‘I’m sorry you are upset about this problem,” but to satisfaction with, ‘I’m glad you are satisfied with our service.'”
Clearly, this kind of tool has big implications for businesses that want to better understand and communicate with customers. When customers are upset or frustrated with their interactions, they may cut off the conversation or even cancel their subscription, membership, or business with that vendor. The tool is currently being used to help engagement in areas such as professional learning, development training, and e-commerce.
While this type of tool could potentially go a long way in improving the customer experience, it’s also important to be mindful that, when training on data sets online, machine learning can sometimes pick up bad behavior. Businesses should remember Microsoft’s Tay—a chatbot meant to be teenage girl, who began spewing racist and sexist remarks less than 24 hour after she emerged on Twitter—as a cautionary tale.
The 3 big takeaways for TechRepublic readers
- A new Tone Analyzer for Customer Engagement for IBM Watson attempts to better detect subtleties in tone to help improve customer service.
- The tool can pick up on seven different types of tone: frustration, satisfaction, excitement, politeness, impoliteness, sadness and sympathy.
- The feature also detects tone throughout the conversation, noting how it changes and giving feedback to agents about when they should become more sympathetic, polite, or excited to help smooth out the interaction.