About big clone bench

[raven4752]

Hello, thank you for your great work!
Would you release the preprocessing steps for code snippets in big clone bench?

[preesee]

Hi I am also researching on dataset bigcloneBench. However, I don't understand how you did the evaluation so that the F1 score is almost 1 in your paper, do you make any negative samples for each clone-type or just classify the type-1,type-2type-3type-4 with the softmax? Any explanation will be helpful ， thanks

[raven4752]

from the code and the paper, I guess the high f1 score comes from dominating positive class - the type-3/type-4 clones.

[preesee]

@raven4752 Sorry to say from my limited research, I am afraid it is really hard to reproduce the score. The other solid paper indicates they only detect no greater than 1% Type3/Type-4 clones
Name |T1 (35,802) |T2 (4,577) |VST3 (4,156) |ST3 (15,031) |MT3 (80,023)| WT3/T4 (7,804,868)
Oreo| 100 |35,798| 99%| 4,547 |100 4,139 89| 13,391 30%| 23,834|0.7 %|57,273
whereas if the dataset is shrinked, it could be happened. The matrix representation is really novel and innovative.