Infer dataset text_raw truncation in data_utils_for_inferring.py

[yangheng95]

Infer dataset text_raw truncation in data_utils_for_inferring.py

[yangheng95]

Thanks for reporting this, the text_right is not used in the BERT-based models, it was used in some models relying on left and right contexts. And this inferring script has not been tested on those non-BERT-based models. I will fix some problems later.

The truncating of sentences is useful in low resources systems, If you use large max_len, it may help, however not very apparent.

Actually, it is difficult to integrate multilingual datasets and multiple models, I have not much time to test new code. Again, sorry for causing your inconvenience.

[yangheng95]

Please check here https://github.com/yangheng95/LC-ABSA/blob/dc27f9b09607e8125c49de9b00cd508a21b9b75a/modules/utils/data_utils_for_inferring.py#L276

[tusharkgerg]

thanks @yangheng95 that was helpful!

[yangheng95]

thanks @yangheng95 that was helpful!

I am sorry that you may check the code again, I update the code because I ignored some cases that aspect is at the beginning or ending.

[tusharkgerg]

@yangheng95 I just ran the code. Seems to have fixed the truncation problem where aspect was in the beginning! :) Awesome Stuff 👍
Also would be great to discuss how I am approaching the problem by using your model implementation.. In case you may have some spare time :p For reassurance.