[ASK] Image-based Dataset Interactions

[preferredai]

[ASK] Image-based Dataset Interactions

[tqtg]

Unfortunately, we don’t keep the original copy of the Amazon dataset.

[tqtg]

(1) The features are original features provided here.

(2) The Amazon Clothing dataset in Cornac is a subset of the original one. The size is smaller because items are filtered to make sure 3 types of modalities available (text, image, and graph). More details can be found here.

(3) Not so clear about your question. Maybe you can define a threshold (rating >= 4.0 as positive) and convert them into binary. Then you can consider them as implicit feedback. In addition, other types of feedback also available in the original dataset (views, clicks, purchases).

[RuihongQiu]

Thank you for your explanation.

So basically, if I want to only use the visual feature of Amazon Clothing, I can enlarge the dataset according to the original dataset since I only use the already extracted features from their original website.

But I can't download the visual feature from their 2014 version dataset, which always points to the 2019 version. And in the 2019 version, there is no pre-extracted visual features.

Could you suggest how to find it?

[tqtg]

I guess you can email the authors and ask for access to the data.

[RuihongQiu]

Thank you.

[RuihongQiu]

Actually, could you guys kindly provide the purely image-based Amazon Clothing dataset, since many recent papers use this dataset along with the Tradesy dataset. It would be much more convenient if we could have a similar dataset. :)

Thank you.