search

KZ-Shi / AMR-CD

Compressing Long Context for Enhancing RAG with AMR-based Concept Distillation
https://arxiv.org/abs/2405.03085#/
1 stars 0 forks source link

Clarity on the Use of Datasets #1

Open RidhiChhajer opened 2 months ago

RidhiChhajer commented 2 months ago

Hello 👋,

I have a question regarding the AMR (Abstract Meaning Representation) graph generation process. Could you please clarify which field in the dataset is responsible for generating the AMR graph for a specific question/row (for both POPQA and Entity-Question dataset)? Where are these long context in the POPQA and Entity-Question datasets ?

Thank you!

KZ-Shi commented 1 month ago

Hi RidhiChhajer,

Thank you for your question, and thanks for your interest in our work.

There is an item in the dataset called "ctxs", representing the retrievaled context corresponding to the question. The context of this item will be converted to the AMR graph.

Kind regards