Generative Q&A example isn't generative

[ssooffiiaannee]

I tried the code in this tutorial for generative q&a, and I find no difference between extractive and generative q&a. The example just extracts the answer and does not construct answers by itself.

Query: when is the next deadpool movie being released
Answers:
[{'answer': ' september 22, 2017'}]

How can I construct a model that answers : "The next deadpool movie will be released on september 22nd, 2017" instead of just extracting the answer from the text.

[JasperLS]

Yes these models seem to stick to provided context. You may be able to get better results trying a LFQA pipeline. Alternatively, the Haystack PromptNode provides a lot of flexibility, e.g., you could use the OpenAI API (if you sign up, you will get some free budget) which generates very good results. I sketched a rough Notebook for you here.