- We will use Task 3 of Diagnostic Questions competition, which was one of the NeurIPS 2020 Competitions.
- The data were collected and kindly shared by Eedi, University of Cambridge, Rice University, Microsoft Research.
- Download the dataset from here or clone a backup repo.
- subfolder structure
images
: actual questionsmetadata
: some side information about questions and studentstrain_data
: usetrain_task_3_4.csv
for trainingtest_data
: usequality_response_remapped_public.csv
for validation andquality_response_remapped_private.csv
for test
- estimate the quality of questions from students responses and other information about questions.
- calculate the rankings of the questions using the template in
submission/template.csv
. - save the file as
YOUR_STUDENT_ID.csv
and submit with your report.
- Report must be a pdf file with filename
YOUR_STUDENT_ID.pdf
- Summary (maximum 250 words): provide a brief overview, the problem statement, methodology, findings, and conclusions
- Introduction: provide background, define the problem, and state your objectives
- Methods: provide technical details (preproprocessing data, models used, details of experiments, how to analyze the results). Include the link to your
GitHub repository
. If you want to keep yor repository private, invitessuai
as a collaborator. - Results: present findings (Use graphs or tables!)
- Discussion: interpret your findings and discuss implications
- Conclusion: summarize main findings, provide the main message, discuss future directions
- References: list of all the sources cited in the report