{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7301572a-4803-4a16-b262-74e41e25803e",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "\n",
    "import pandas as pd\n",
    "from datasets import Dataset, DatasetDict, load_dataset\n",
    "from huggingface_hub import Repository, create_repo\n",
    "from selectolax.parser import HTMLParser"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a898baed-3640-4ca8-9d3d-b88f6c85a428",
   "metadata": {},
   "outputs": [],
   "source": [
    "def _parse_start_end(node):\n",
    "    return int(node.attrs[\"start\"][1:]), int(node.attrs[\"end\"][1:])\n",
    "\n",
    "\n",
    "def get_original_text(sent_toks) -> str:\n",
    "    empty_tokens = [i for i, t in enumerate(sent_toks) if not t.text().strip()]\n",
    "    org_sent_toks = [t.text() for i, t in enumerate(sent_toks) if not i in empty_tokens]\n",
    "    return \" \".join(org_sent_toks)\n",
    "\n",
    "\n",
    "def get_corrected_text(toks_cor, last_end, sent_end) -> str:\n",
    "    cor_toks = []\n",
    "    for tok in toks_cor:\n",
    "        tok_start, tok_end = _parse_start_end(tok)\n",
    "        if tok_start >= last_end and tok_end <= sent_end:\n",
    "            cor_toks.append(tok.text())\n",
    "            last_end = tok_end\n",
    "    return last_end, \" \".join(cor_toks)\n",
    "\n",
    "\n",
    "def process_doc(doc, path):\n",
    "    toks = doc.select('tier[category=\"tok\"] event').matches\n",
    "    toks_cor = doc.select('tier[category=\"TH1\"] event').matches\n",
    "    sents = doc.select('tier[category=\"sentence\"] event').matches\n",
    "\n",
    "    last_end = 0\n",
    "    for sent_no, org_sent in enumerate(sents):\n",
    "        sent_start, sent_end = _parse_start_end(org_sent)\n",
    "        sent_toks = toks[sent_start:sent_end]\n",
    "        original_text = get_original_text(sent_toks)\n",
    "        last_end, corrected_text = get_corrected_text(toks_cor, last_end, sent_end)\n",
    "\n",
    "        yield (\n",
    "            {\n",
    "                \"original\": original_text,\n",
    "                \"corrected\": corrected_text,\n",
    "                \"id\": f\"{path.stem}-{sent_no}\",\n",
    "            }\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "191fe2e3-8a4e-47e2-9316-5b6028662c02",
   "metadata": {},
   "outputs": [],
   "source": [
    "DATASET_NAME = \"merlin\"\n",
    "dataset_path = Path.home() / DATASET_NAME\n",
    "if not Path(dataset_path).exists():\n",
    "    repo_url = create_repo(name=DATASET_NAME, repo_type=\"dataset\")\n",
    "    repo = Repository(local_dir=str(dataset_path), clone_from=repo_url)\n",
    "    repo.lfs_track(\"*.jsonl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "005d059f-5de4-4f14-bde3-0cc9ff2435c0",
   "metadata": {},
   "outputs": [],
   "source": [
    "MERLN_EXMARALDA_BASE = Path.home() / Path(\n",
    "    \"Downloads/MERLIN Written Learner Corpus for Czech, German, Italian 1.1/merlin-exmaralda-v1.1/\"\n",
    ")\n",
    "\n",
    "for lang in (\"german\", \"czech\", \"italian\"):\n",
    "    lang_docs = []\n",
    "    for path in (MERLN_EXMARALDA_BASE / lang).glob(\"*.exb\"):\n",
    "        with open(path) as fp:\n",
    "            xml = HTMLParser(fp.read())\n",
    "            docs = list(process_doc(xml, path))\n",
    "            lang_docs.extend(docs)\n",
    "    Dataset.from_dict(pd.DataFrame(lang_docs)).to_json(dataset_path / f\"{lang}.jsonl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f9d396b2-98dc-4c04-950f-0332a3a6d751",
   "metadata": {},
   "outputs": [],
   "source": [
    "repo.push_to_hub()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "91377eaf-ffac-4df0-9c85-4f6ee979f99f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}