| import os | |
| import json | |
| import hashlib | |
| data = json.loads(open("dev.json").read()) | |
| with open("queries.jsonl", "w") as f: | |
| for d in data: | |
| q = {"_id": d["_id"], "text": d["question"], "metadata": {"answer": [d["answer"]]}} | |
| f.write(json.dumps(q) + "\n") | |
| with open("corpus.jsonl", "w") as f: | |
| for d in data: | |
| for idx, ci in enumerate(d["context"]): | |
| c = {"_id": f"{d['_id']}_{idx}", "text": ci[0] + ": " + " ".join(ci[1])} | |
| f.write(json.dumps(c) + "\n") | |
| here = os.path.dirname(os.path.abspath(".")) | |
| def prepare_json_dataset(): | |
| hash_ids = set() | |
| with open(os.path.join(here, "corpus.txt"), "w") as fout: | |
| with open(os.path.join(here, "corpus.jsonl"), "r") as fin: | |
| for line in fin: | |
| line = json.loads(line) | |
| hash_id = hashlib.md5(line["text"].encode("utf-8")).hexdigest() | |
| if hash_id in hash_ids: | |
| continue | |
| else: | |
| fout.write(line["text"] + "\n") | |
| hash_ids.add(hash_id) | |
| prepare_json_dataset() | |