About the data in paper

[zhaojingj...@gmail.com]

I notice that there are several results on the test sets from the three previous editions of this subtask in paper SemEval-2016 Task 4: Sentiment Analysis in Twitter. But we didn't run our system on these sets, even when I train my system, I used  these data. So,

Should we train our system again without these test set?? If so ,please let us know, it takes sometime after all.

And  since I just have the result over test set 2017, should I just put this result into my paper?? Or,you know, train my system again and test over several test set ,then put them into my paper??