Question-answer-Generation-from-Chest-X-ray-Reports

Important files:

• qg.py, statistics.py, qa.py, utils.py
• reports.txt
• keys, reports (directory)
• By the way, the directory reports include 7430 files, which is hard to upload all. reports.txt includes all the information, if you want to directly run qg.py, first recover all the reports files according to reports.txt

1. question-answer generation according to pre-defined key words in directory keys

   python3 qg.py

   • this process first simplify the original reports
   • results are saved in a new directory: results
   • main results include: simplified_reports.txt, patterns (i.e., the coding files), all_answers.txt, and all_questions.txt
   • printed information:

        number of reports_files:  7430
        Number of yeshow_questions: 136
        Number of yes_questions: 28700
        number of in_questions: 2271
        number of within_questions: 4503
        number of upper_questions: 270
        number of status_rl_questions: 87
        Number of slq_questions: 26278
        Number of status_questions: 13
        Number of location_questions: 2948
        number of in_questions: 8
        All original generated questions:  65214
        Number of question_answer pairs after post-process:  63232
   

2. statistics on the above generated question-answer pairs

   python3 statistics.py

   • results are saved in a new directory: qg-statistics
   • main results include: select_questions.txt

3. question answering to augment the QA pairs

   python3 qa.py

   • results are saved in a new directory: qa

   • main results include: qa_answers.txt, qa_questions.txt

   • main results include: qa_answers.txt, qa_questions.txt

4. This work is accepted as a full paper by the 32nd IEEE CBMS International Symposium on Computer-Based Medical Systems.