yusenzhang826/aim

README

Codes and data for the Coling2022 paper:
AiM:Taking Answers in Mind to Correct Chinese Cloze Tests in Educational Applications

The datasets are available now and codes will come soon.

The EinkCC and SynCC dataset are available.
The meanings of the fields in the json file are as follows:
1.file: the image path
2.handwriting: the hand-written content in the image
3.answer: the true answer for the question
4.bin_label(if exists): the binary sequence label for the answer, which indicates whether the token in the true answer appears in the hand-written content. Notice that the binary sequence labels are not used in practice.
5.edit_label(if exists): the edit sequence label for the answer. For details please refer to the paper.
6.label: The label indicates whether the hand-written content is a correct anwser(1 for correct and 2 for wrong)

For the HWCC, you can get the HWDB2.x benchmark first. Then you can generate HWCC dataset for the CCC task through the function random_extend_data in the random_generate_neg_data.py script.