Sentiment analysis or opinion mining as it’s sometimes called can be a very powerful tool when creating a more engaging experience for your chatbot users. One of the biggest problems that we encounter when measuring satisfaction scores in chatbots is that users report frustration when communicating with a bot. The problem is that the chatbot doesn’t know that a user is frustrated and continues business as usual which amplifies the problem.
With sentiment analysis your chatbot can tailor the dialog based on the sentiment score calculated from the user input. This sentiment score can tell the bot if the user is having a positive, neutral or a negative experience. Here are some interesting ways that you can use this information to improve your chatbot:
- Suggest human handoff. Measure the sentiment score for each interaction and set a threshold score that will initiate a human handoff. For example, if the bot notices that the score is reduced with each interaction it can gently suggest something like the following: “I’m sorry I’m not doing a good job helping you today would you like me to transfer you to one of my human teammates?”
- Customer scoring. You can measure the sentiment score average for all conversation over a given time period. Based on this you can score customers and decide on a course of action. One example of this would be being able to contact a customer with a negative sentiment towards your message before they go to the Internet and write a bad review.
- Marketing campaign analytics. Use the sentiment score to determine if a marketing campaign is well received by customers. With traditional marketing analytics tools, you get raw numbers that don’t tell the full story. With sentiment analysis, you can measure the emotional response that clients experience when seeing a given marketing campaign.
- Identify brand ambassadors. The people with the highest sentiment scores are most likely to be your brand ambassadors. This can help you identify those individuals and help them be more successful as your brand ambassadors. For example, you can send special promotions to these individuals to further encourage them to represent your brand in the best light possible.
This is barely scratching the surface of what is possible with sentiment analysis.
Can you contribute any ideas on using sentiment analysis to improve your chatbot?