data_preparation.py

""" Script to prepare the SQuAD2.0 data to the GEM format

    @author: AbinayaM02
"""

# Import libraries
import json
import pandas as pd
from sklearn.model_selection import train_test_split

# Function to generate gem id
def add_gem_id(data: dict, split: str) -> dict:
    """ 
    Add gem id for each of the datapoint in the dataset.

    Parameters:
    -----------
    data: dict,
        data.
    split: str,
        split of data (train, test or validation).

    Returns:
    --------
    dict
        dictionary with updated id
    """
    gem_id = -1
    generated_data = {"data": []}
    for example in data:
        temp_dict = {}
        title = example["title"]
        for paragraph in example["paragraphs"]:
            context = paragraph["context"]  # do not strip leading blank spaces GH-2585
            for qa in paragraph["qas"]:
                question = qa["question"]
                qa_id = qa["id"]
                answer_starts = [answer["answer_start"] for answer in qa["answers"]]
                answers = [answer["text"] for answer in qa["answers"]]
                # Features currently used are "context", "question", and "answers".
                # Others are extracted here for the ease of future expansions.
                gem_id += 1
                temp_dict["id"] = qa_id
                temp_dict["gem_id"] = f"gem-squad_v2-{split}-{gem_id}"
                temp_dict["title"] = title
                temp_dict["context"] = context
                temp_dict["question"] = question
                temp_dict["answers"] = {
                        "answer_start": answer_starts,
                        "text": answers,
                    }
                generated_data["data"].append(temp_dict)
    return generated_data


# Function to split data
def split_data(file_name: str, data_type: str) -> (dict, dict):
    """
    Method to split the data specific to SQuAD2.0

    Parameters:
    -----------
    file_name: str,
        name of the file.
    data_type: str,
        type of the data file.

    Returns:
    --------
    (dict, dict)
        split of data
    """
    
    if data_type == "json":
        with open(file_name, 'r') as json_file:
            data = json.load(json_file)["data"]
            json_file.close()
    
    # split the data into train and test
    train, test = train_test_split(data, train_size=0.7, random_state = 42)
    return(train, test)


if __name__ == "__main__":
    # split the train data
    train, test = split_data("squad_data/train-v2.0.json", "json")
    
    # add gem id and save the files
    train = add_gem_id(train, "train")
    test = add_gem_id(test, "test")
    
    # save the train split
    with open("train.json", "w") as train_file:
        json.dump(train, train_file, indent = 2)
        train_file.close()

    # save the test split
    with open("test.json", "w") as test_file:
        json.dump(test, test_file, indent = 2)
        test_file.close()
    
    # load validation data
    with open("squad_data/dev-v2.0.json", "r") as dev_file:
        validation = json.load(dev_file)["data"]
        dev_file.close()
    
    # add gem id and save valid.json
    validation = add_gem_id(validation, "validation")
    with open("valid.json", "w") as val_file:
        json.dump(validation, val_file, indent = 2)
        val_file.close()