utils/convert_hgf_to_ir_format.py

"""
This script only convert huggingface dataset into topic, qrels and folds files.
To prepare corpus falls in the JsonCollection in Anserini/Pyserini format, refer to the scripts in pyserini repository. 
"""

import os
import sys
import json
import gzip
from glob import glob

from tqdm import tqdm
from datasets import load_dataset

dirname, abspath, os_join = os.path.dirname, os.path.abspath, os.path.join
dataset_dir = dirname(dirname(abspath(__file__)))
# output_dir = os_join(dataset_dir, "ir-format-dataset")
output_dir = dataset_dir

version = "v1.1"
languages = [
    'arabic',
    'bengali',
    'english',
    'indonesian',
    'finnish',
    'korean',
    'russian',
    'swahili',
    'telugu',
    'thai',
    'japanese',
]


def get_output_dir(lang):
    lang_dir = os_join(output_dir, f"mrtydi-{version}-{lang}", "ir-format-data")
    os.makedirs(lang_dir, exist_ok=True)
    return lang_dir


def write_topic(qid2query, outp_fn):
    with open(outp_fn, "w") as f:
        for qid in qid2query:
            f.write(f"{qid}\t{qid2query[qid]}\n")
 

def write_qrels(qid2reldocs, outp_fn):
    with open(outp_fn, "w") as f:
        for qid in qid2reldocs:
            for docid in qid2reldocs[qid]:
                f.write(f"{qid}\tQ0\t{docid}\t1\n")


def convert_set(single_set):
    qid2query = {}
    qid2reldocs = {}
    for data_entry in single_set:
        qid = data_entry["query_id"]
        query = data_entry["query"]
        pos_docids = [doc["docid"] for doc in data_entry["positive_passages"]]
        qid2query[qid] = query
        qid2reldocs[qid] = pos_docids
    return qid2query, qid2reldocs


def prepare_language(lang):
    dataset = load_dataset('utils/local-mr-tydi.py', lang)

    set_names = ["train", "dev", "test"]
    folds = {}
    output_dir = get_output_dir(lang)
    for set_name in set_names:
        topic_fn = os_join(output_dir, f"topics.{set_name}.txt")
        qrel_fn = os_join(output_dir, f"qrels.{set_name}.txt")

        qid2query, qid2reldocs = convert_set(dataset[set_name])
        write_topic(qid2query, topic_fn)
        write_qrels(qid2reldocs, qrel_fn)


def main():
    for lang in languages:
        prepare_language(lang)


if __name__ == "__main__":
    main()