Playing around with Retrieval Augmented Generation.
- Go through first tutorial
- Try this with my own dataset (e.g expenses csv)
- Repeat with my own dataset but this time:
- See if there's a converter for CSV to Haystack Document
- Split and clean the document
- Then see if it provides better answers
- Understand the pipeline connections (draw a diagram)
- Make it interactive, with the abiliy to add documents and add questions about newly added documents
- What are the different configurations and what do they mean?
## Questions
- In Lesson 2, Why did we initialise the InMemoryDocumentStore with
BM25Plusand useInMemoryBM25Retriever?
Actually, Hugging Face gave a funny answer in lesson 3. Just noting it here.
** Which of the following recipes would be meet with the most disapproval from Ron Swanson? Explain why in Ron Swanson's style of speech**
Ron Swanson would most likely disapprove of the vegan persimmon flan recipe due to its use of non-traditional ingredients such as persimmon pulp, agar agar, and coconut creme.
In Ron Swanson's style of speech, he might say, "This flan is an abomination! Persimmon pulp in a flan? That's like putting a turkey in a lasagna! It's an affront to all that is holy and delicious in this world.
I'll have my men round up every copy of this recipe and burn it in a ceremonial bonfire. From this day forward, flan will be made the way it was meant to be made: with eggs, sugar, and a healthy dose of manliness."