{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "003ecb55",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "from typing import Dict, List\n",
    "from tqdm import tqdm\n",
    "\n",
    "import pandas as pd\n",
    "import srsly\n",
    "import spacy\n",
    "from spacy.training.iob_utils import biluo_to_iob\n",
    "from spacy.training import offsets_to_biluo_tags\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d43fd200",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset_path = Path(\"promises-elections-pl-twitter-1-reviewed-fixed-case.jsonl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f2ffadfc",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = list(srsly.read_jsonl(dataset_path.as_posix()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e89bde09",
   "metadata": {},
   "outputs": [],
   "source": [
    "import uuid\n",
    "from IPython.display import display_javascript, display_html\n",
    "import json\n",
    "\n",
    "class RenderJSON(object):\n",
    "    def __init__(self, json_data):\n",
    "        if isinstance(json_data, dict):\n",
    "            self.json_str = json.dumps(json_data)\n",
    "        else:\n",
    "            self.json_str = json\n",
    "        self.uuid = str(uuid.uuid4())\n",
    "        \n",
    "    def _ipython_display_(self):\n",
    "        display_html('<div id=\"{}\" style=\"height: 600px; width:100%;\"></div>'.format(self.uuid),\n",
    "            raw=True\n",
    "        )\n",
    "        display_javascript(\"\"\"\n",
    "        require([\"https://rawgit.com/caldwell/renderjson/master/renderjson.js\"], function() {\n",
    "          document.getElementById('%s').appendChild(renderjson(%s))\n",
    "        });\n",
    "        \"\"\" % (self.uuid, self.json_str), raw=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fed644a9",
   "metadata": {},
   "outputs": [],
   "source": [
    "RenderJSON(data[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4fad229b",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df = pd.DataFrame(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9f75c8b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "data[0][\"text\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "eadb23f1",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df = data_df[data_df.answer == \"accept\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "261e6a8c",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df.text[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cbe9e877",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df.text[1][202:224]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5583aa66",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_df.spans[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cedd7c42",
   "metadata": {},
   "outputs": [],
   "source": [
    "def convert_spacy_to_iob(text: str, spans: List[Dict], nlp: spacy.language.Language) -> str:\n",
    "    doc = nlp(text)\n",
    "    entities = [(span[\"start\"], span[\"end\"], span[\"label\"].upper()) for span in spans]\n",
    "    if entities:\n",
    "        main_entity = list(set(span[\"label\"].upper() for span in spans))[0]\n",
    "    else:\n",
    "        main_entity = None\n",
    "    tags = offsets_to_biluo_tags(doc, entities)\n",
    "    # then convert L->I and U->B to have IOB tags for the tokens in the doc\n",
    "    # tags_corrected = [tag.replace(\"L-\", \"I-\").replace(\"U-\", \"B-\") for tag in tags]\n",
    "    tags_corrected = biluo_to_iob(tags)\n",
    "    tokens = [token.text for token in doc]\n",
    "    return {\n",
    "        \"tokens\": tokens, \n",
    "        \"tags\": tags_corrected,\n",
    "        \"label\": main_entity or \"\"\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c7d3aa1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from spacy.lang import pl\n",
    "\n",
    "nlp = pl.Language()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a0449216",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pprint import pprint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4edc0665",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokens, tags, label = convert_spacy_to_iob(data_df.text[0], data_df.spans[0], nlp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "db5c6b07",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dee2cf4f",
   "metadata": {},
   "outputs": [],
   "source": [
    "tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "656c39a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "37e46df4",
   "metadata": {},
   "outputs": [],
   "source": [
    "iob_dataset = pd.DataFrame([\n",
    "    {\n",
    "        **convert_spacy_to_iob(row[\"text\"], row[\"spans\"], nlp),\n",
    "        # \"url\": row[\"url\"],\n",
    "        # \"tweet_id\": row[\"url\"].split(\"/\")[-1],\n",
    "    }\n",
    "    for _, row\n",
    "    in tqdm(data_df.iterrows())\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "74c3d161",
   "metadata": {},
   "outputs": [],
   "source": [
    "iob_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0caa005a",
   "metadata": {},
   "outputs": [],
   "source": [
    "iob_dataset.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d121b8fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "from more_itertools import flatten "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "55212d38",
   "metadata": {},
   "outputs": [],
   "source": [
    "sorted(set(flatten(iob_dataset.tags)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8063ab0f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9787243d",
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = train_test_split(iob_dataset, test_size=0.4)\n",
    "dev, test = train_test_split(test, test_size=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bcb323fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "train.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f11f7de6",
   "metadata": {},
   "outputs": [],
   "source": [
    "train.reset_index(inplace=True, drop=True)\n",
    "train.drop(columns=[\"label\"], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "459dad3c",
   "metadata": {},
   "outputs": [],
   "source": [
    "train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9bdcab49",
   "metadata": {},
   "outputs": [],
   "source": [
    "len(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aeff2557",
   "metadata": {},
   "outputs": [],
   "source": [
    "len(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d1301717",
   "metadata": {},
   "outputs": [],
   "source": [
    "len(dev)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d55c9edc",
   "metadata": {},
   "outputs": [],
   "source": [
    "test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "636e9268",
   "metadata": {},
   "outputs": [],
   "source": [
    "test.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f5486e11",
   "metadata": {},
   "outputs": [],
   "source": [
    "test.reset_index(inplace=True, drop=True)\n",
    "test.drop(columns=[\"label\"], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "34aa63ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c1bc32c4",
   "metadata": {},
   "outputs": [],
   "source": [
    "dev"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c8ad5315",
   "metadata": {},
   "outputs": [],
   "source": [
    "dev.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5591dea4",
   "metadata": {},
   "outputs": [],
   "source": [
    "dev.reset_index(inplace=True, drop=True)\n",
    "dev.drop(columns=[\"label\"], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "53f0c791",
   "metadata": {},
   "outputs": [],
   "source": [
    "dev"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fdb29857",
   "metadata": {},
   "outputs": [],
   "source": [
    "train.to_parquet(\"../train.parquet\")\n",
    "test.to_parquet(\"../test.parquet\")\n",
    "dev.to_parquet(\"../dev.parquet\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c5b20f97",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_parquet(\"../train.parquet\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e05e0d83",
   "metadata": {},
   "outputs": [],
   "source": [
    "df"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "97bd981c6355438647fbd6d64b9445f9e50a1f8ddfec47bd063c9ea6e8fe3e87"
  },
  "kernelspec": {
   "display_name": "Python 3.9.5 ('embeddings')",
   "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": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}