qa_pvalue.py

import json
from tqdm import tqdm
import scipy.stats as stats
def evaluate_retrieval(retrieval_file, topk):
    retrieval = json.load(open(retrieval_file))
    accuracy = { k : [] for k in topk }
    max_k = max(topk)
    for qid in tqdm(list(retrieval.keys())):
        contexts = retrieval[qid]['contexts']
        has_ans_idx = max_k  # first index in contexts that has answers

        for idx, ctx in enumerate(contexts):
            if idx >= max_k:
                break

            if ctx['has_answer']:
                has_ans_idx = idx
                break

        for k in topk:
            accuracy[k].append(0 if has_ans_idx >= k else 1)

    return accuracy

if __name__=="__main__":
    topk = [5, 10, 20, 50, 100]
    all_scores_a,all_scores_b=None,None
    for DATA in ["nq","trivia","wq","curated","squad"]:
        FileNameA="/data/t-junhanyang/InfoCSE/QA_TEST/InfoCSE_ICT_K1.{}.test.json".format(DATA)
        FileNameB="/data/t-junhanyang/InfoCSE/QA_TEST/CONPONO.{}.test.json".format(DATA)
        scores_a=evaluate_retrieval(FileNameA,topk)
        if all_scores_a is None:
            all_scores_a=scores_a
        else:
            for k in topk:
                all_scores_a[k]+=scores_a[k]
        print(FileNameB)
        scores_b=evaluate_retrieval(FileNameB,topk)
        if all_scores_b is None:
            all_scores_b=scores_b
        else:
            for k in topk:
                all_scores_b[k]+=scores_b[k]
    print(len(all_scores_a[5]))
    print(len(all_scores_a[100]))
    for k in topk:
        stat_val, p_val = stats.ttest_ind(all_scores_a[k], all_scores_b[k])
        print(str(k)+': '+str(p_val/2))