How can I get the dataset?

[hxiaoj]

how can I get the dataset? I have download the dataset from eedi but i don't know how to process it

[alexscarlatos]

Hi! Apologies for the late response. The data files (.csv) should have the following columns:

• id: the unique question ID
• question: the question text
• correct_option: a JSON object containing:
  • option: the correct option text
  • explanation: the textual solution for the question
• construct_info: a JSON object containing:
  • construct1: a list for the top-level construct: [construct ID: int, construct description: str]
  • construct2: a list for the mid-level construct: [construct ID: int, construct description: str]
  • construct3: a list for the low-level construct: [construct ID: int, construct description: str]
• distractors: a JSON list of 3 objects (one for each distractor), each containing:
  • option_idx: the index of the option in the question (1-4)
  • option: the distractor option text
  • explanation: the feedback/explanation text for the distractor
  • proportion: the percentage of students that chose this option (set to 0 if unknown)
  • misconception: the text for the misconception for this distractor (optional, only needed for misconception-based ICL prompts or rule-based method)

You can specify the train and test file paths using the data.trainFilepath and data.testFilepath command line arguments. Hope that helps!