Question about data preprocessing #22
Description
I have a question regarding the item text encoding: in the paper, only Title, Brand, Categories, and Price are used as item textual features. However, the original Amazon dataset also contains a "description" field, which is longer and more informative. Could you share the reasoning behind not using the description field? Is it mainly due to efficiency concerns, or does including it not improve performance significantly?
Additionally, I would like to experiment with including the description in the item textual representation for semantic encoding. If possible, would you be willing to share the data preprocessing code or provide guidance on how to include it?