In some ML applications, algorithmic bias can creep in — and it can quickly lead to unexpected outcomes. But how does it happen, and how can we avoid this?
Introducing Training Optimizer: A New Way to Improve Your Bot's Understanding Ability
Today, we are excited to announce the launch of the Training Optimizer. This new feature improves your shopping bot’s ability to understand customer questions.
Why did we build the Training Optimizer?
We built the Training Optimizer for two reasons.
Training Optimizer eliminates the need for the chatbot to seek clarifications. Every day, our Machine Learning team focuses on improving the recognition rate of our client’s AI chatbot (recognition rate is defined as the percentage of questions your bot is able to understand and answer). In doing so, they identified a core problem that leads to a lower recognition rate. They found that as the number of Training Questions increases, so does the likelihood of some questions overlapping with one another. This prompts the bot to seek clarification from customers before it can serve them the right Answer. Training Optimizer solves this problem.
Secondly, we know that managing your bot’s content and identifying similar Training Questions can be time-consuming. We decided to automate this for you through the Training Optimizer.
How does it work?
Training Optimizer will check daily for similar training across different Answers in your conversational AI. It looks at several conditions, including how often customers historically have interacted with these Training Questions, and then provides a ranked and prioritized list. With a few clicks, you can easily take action to resolve these overlapping similarities between Training Questions.
Our design team built a creative two-column view solution which makes it easy to move Training Questions between the Answers. You can also edit or delete a Training Question if required.
After you have fixed the identified list of training similarities, click the update button. That's it!
For more details on how to use the Training Optimizer, please see the documentation in the Ada Resource Centre.
How can you measure success?
To determine if your changes are making a difference, keep an eye on your bot's clarification rate. You can find this report in the Analytics section. As a refresher, the Clarification rate is the percentage of conversations in which the bot responded with the Needs Clarification Answer at least once.
We believe that Training Optimizer will greatly improve your customer's experience with your bot in a meaningful way.
As always, if you have any feedback or ideas about improvement, please reach out to your Account Manager to let us know. We are constantly looking for ways to improve our product.
To learn more about optimizing your customer support strategy, check out this article from our ACX team on the 7 Keys for Optimizing a CX Chatbot.