Is anyone using Embedded Policy only and found it performing well in production?
Hey @sibbsnb. Looking forward to get more community responses on that too. Are you having any issues with it?
May I know if you are talking about the supervised and unsupervised embedding in config file.
It really depends on what you want to achieve… I am not that skilled in understanding all the in-depth NLP processing. Generally, I added the unpacked supervised embedding with some custom component. It works well for my project.
I have about 150 intents with 2k of examples, I am still trying to perfect it but supervised learning is needed for my training data, hence, I use the components under Supervised Embedding…
I often use Rasa test to check the performance of each intent, then improve it later.
Tried with few, looks ok. Wanted to see if anyone tried it in detail in prod.
I was referring to the dialog management policy not the NLP