For measuring recall, does the word ordering matter? If both the retrieved text and highlights are quite different semantically, the recall might be descent if they both contain very common words, although this may not happen often. Have you considered using ROUGE-L instead? It seems to be a good candidate for measuring recall here.
This couldn't have been more timely and thorough. Instantly liked and subscribed :) You deserve so much more attention.
Hope it helped!
For measuring recall, does the word ordering matter? If both the retrieved text and highlights are quite different semantically, the recall might be descent if they both contain very common words, although this may not happen often. Have you considered using ROUGE-L instead? It seems to be a good candidate for measuring recall here.
Thank you!! This is gem 💎
Cheers!
what i understand is that chunking strategies only apply to traditional rag, but not when using KG (via triples)... right?
Still need to chunk data into nodes/relationships I believe.
Congratulations, this content was perfect.
Preciate it!