Building a Q&A bot using RASA NLU component

[Deprecated] Rasa X Community Edition
1

Hi, I am trying to build a Q&A bot where the data is a CSV (one column with Question and the other is an answer) In this case, would RASA help as there is no context needed or a conversation flow.

image
image782×461 15.6 KB

So, in the above pic, I was trying an example where the responses are hard-coded. For “Corona_intro” intent, the response will be “utter_corona_intro”. I did see few other posts here saying NLU core is not needed for this as there not really a custom script or actions needed for this.

So this is just a workaround, but what if my Q&A pairs are longer length. Is there a feature/add on in RASA that helps me to import, say a CSV and convert those Q&A to components in the domain.yml file and train the model?

Also, not much complex stories would be needed in this scenario as it is a direct Q&A pair.

Hi @rctatman, As discussed in today’s live coding session, it would be great if you could help with the feature that helps achieve this use case.

Thanks

2

Hi @vishnuragas,

Ideally for Q&A/FAQ bot the desired approach would be to use response selector . Your stories file will look lot clean and simple because of using response selector as there is no impact on flow. Further, the domain file will look even smaller considering the mappings recorded in a different mark down file.

1 Like
3

Thanks for jumping in, sandilya45! A response selector & retrieval action was exactly what I was thinking of. :smile: (This blog post goes into more detail.)

@vishnuragas: the specific thing I was half-remembering was that there was a recent-ish pull request merged that added the full retrieval intent name to the tracker.