import json import os from itertools import combinations from random import seed, randint, shuffle import pandas as pd from datasets import load_dataset def get_stats(filename): with open(filename) as f: _data = [json.loads(i) for i in f.read().splitlines()] return len(_data), list(set([len(i['choice']) for i in _data])), len(list(set([i['prefix'] for i in _data]))) def create_analogy(_data): analogy_data = [] seed(12) for i in _data: source = [] target = [] for s, t in zip(i['source'], i['target']): if s not in source and t not in target: source.append(s) target.append(t) assert len(source) == len(target), f"{len(source)} != {len(target)}" all_combinations = list(combinations(range(len(source)), 2)) for n, (q_h_id, q_t_id) in enumerate(all_combinations): choice = [[target[x], target[y]] for m, (x, y) in enumerate(all_combinations) if m != n] answer_id = randint(0, len(source) - 1) choice = choice[:answer_id] + [[target[q_h_id], target[q_t_id]]] + choice[answer_id:] assert choice[answer_id] == [target[q_h_id], target[q_t_id]] analogy_data.append({ "stem": [source[q_h_id], source[q_t_id]], "choice": choice, "answer": answer_id, "prefix": i["type"] }) return analogy_data data = load_dataset("relbert/scientific_and_creative_analogy", split='test') data = create_analogy(data) data_m = [i for i in data if i['prefix'] == 'metaphor'] data_s = [i for i in data if i['prefix'] != 'metaphor'] seed(12) shuffle(data_m) shuffle(data_s) validation = data_s[:int(0.1 * len(data_s))] + data_m[:int(0.1 * len(data_m))] test = data_s[int(0.1 * len(data_s)):] + data_m[int(0.1 * len(data_m)):] os.makedirs("dataset/scan", exist_ok=True) with open("dataset/scan/valid.jsonl", "w") as f: f.write("\n".join([json.dumps(i) for i in validation])) with open("dataset/scan/test.jsonl", "w") as f: f.write("\n".join([json.dumps(i) for i in test])) t_size, t_num_choice, t_relation_type = get_stats("dataset/scan/test.jsonl") v_size, v_num_choice, v_relation_type = get_stats("dataset/scan/valid.jsonl") stat = [{ "name": "`scan`", "Size (valid/test)": f"{v_size}/{t_size}", "Num of choice (valid/test)": f"{','.join([str(n) for n in v_num_choice])}/{','.join([str(n) for n in t_num_choice])}", "Num of relation group (valid/test)": f"{v_relation_type}/{t_relation_type}", "Original Reference": "[relbert/scientific_and_creative_analogy](https://huggingface.co/datasets/relbert/scientific_and_creative_analogy)" }] print(pd.DataFrame(stat).to_markdown(index=False))