How should I leverage the sentimental detection .How do I improve bots accuracy or human behaviour on this?
There is a blog post with an example of how to use sentiment analysis with Rasa here. If you have further questions, please specify what you’ve done, what you are trying to achieve and what exactly your question is.
Yes , I have done sentimental analysis on my bot going through this blog.Currently I used pretrained model .It will detect sentiment just like entity .How can I leverage this detection to make the reply from bot end more human .How do I achieve this in rasa .Via custom actions? or Stories on intent.If via custom action I would want a better picture
As a first step, you’d need to find out what your idea is on a conceptional level for how to use the extracted sentiment information. Only after that, the technical question of how to do this in rasa becomes relevant.
There are many ways how sentiment analysis can be used to improve chatbot interactions. A few are outlined in this blog post: How Chatbots Use Sentiment Analysis to Improve Customer Satisfaction
The technical solution in rasa depends completely on what you choose to do:
- If you e.g. want your bot to respond differently depending on the extracted sentiment entity, you can add stories with those different responses.
- If you want to hand over to a human when the user is e.g. angry, you can find some advice on how to implement human hand-off here.
- If you want anything else to happen based on the extracted sentiment entity (e.g. asking them for their name and email address and adding them to a customer database or whatever else you come up with), write a custom action that does that, and add a story that contains this action following a statement with the respective sentiment entity.
Maybe if someone from the community has an example of how they use sentiment analysis with their bot, they can share their idea here?