ronaldseoh / atsc_prompts Goto Github PK

Hello I am really interested in your paper and especially thank you for opening your code for replication.
I was trying to study your method and in the process, I was stuck in further pre-training BERT with Yelp Review dataset.

At your paper, it was said that you extracted review dataset related to restaurant, and I think it is the "restaurantIDs.txt" if I am right. However, I cannot find the ID text file in your repository, so can I know the source or the method you extracted restaurant IDs??

Thank you for your work!
Best regards,
kyuri.

Hello. Can you provide the file "restaurantIDs.txt"? Thank you.

the error info is:

FAILURE TO GENERATE DATASET !
Found duplicate Key: 0
Keys should be unique and deterministic in nature

I see this code:

            if len(aspect_term_sentiment) > 0:
                for ats in aspect_term_sentiment:
                    yield id_, {
                        "text": sentence_text,
                        "aspect": ats["aspect"],
                        "sentiment": ats["sentiment"]
                    }

In my view, for every aspect in a text, the id_ will must be duplicate, isn't it ?

,I can't find path = '../progress/lm_further_pretraining_bert_amazon_electronics_bseoh_2021-03-20--13_28_15/results/checkpoint-2360776'.