There is a problem in the processed dataset, all aspect terms are mentioned as None in the second line of each review. In the transfer.py file, u select the target tag in the xml file, which doesn't exist in SemEval 2014 files but in 2016 version.
So I think u must use term instead of target
But what I cannot understand why I got very closed results to what u report in this repo even the aspect terms were masked?
Does this make any sense or am I missing something?
Hi,
There is a problem in the processed dataset, all aspect terms are mentioned as None in the second line of each review. In the transfer.py file, u select the target tag in the xml file, which doesn't exist in SemEval 2014 files but in 2016 version. So I think u must use term instead of target
But what I cannot understand why I got very closed results to what u report in this repo even the aspect terms were masked?
Does this make any sense or am I missing something?
Many thanks in advance