{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found data pair: data/Watch Out, Apple... [5v72vj2nFN0].en.txt and data/Watch Out, Apple... [5v72vj2nFN0].cn.txt\n",
      "Found data pair: data/It's Finally Over. [KJF5Nn23CwM]-4k.en.txt and data/It's Finally Over. [KJF5Nn23CwM]-4k.cn.txt\n",
      "Found data pair: data/Apple, Are U Ok？ [34R2xVzBmfA].en.txt and data/Apple, Are U Ok？ [34R2xVzBmfA].cn.txt\n",
      "Found data pair: data/China lost its GPU privileges [Wz9JoTTRa4k].en.txt and data/China lost its GPU privileges [Wz9JoTTRa4k].cn.txt\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "import json\n",
    "\n",
    "INSTRUCT_CHUNKED_PROMPT = \"\"\"你是一个擅长翻译科技新闻的翻译专家。请将以下内容翻译为中文，使用相同格式输出，并保留时间戳。不要漏掉任何信息。合并多行文本时，保留第一个和最后一个时间戳。\"\"\"\n",
    "\n",
    "DOCUMENT_ROOT = \"data\"\n",
    "files = listdir(\"data\")\n",
    "\n",
    "all_datapoints = []\n",
    "\n",
    "for f in files:\n",
    "    en_fname = join(DOCUMENT_ROOT, f)\n",
    "    if en_fname.endswith(\".en.txt\") and isfile(en_fname):\n",
    "        cn_fname = join(DOCUMENT_ROOT, f.replace(\".en.txt\", \".cn.txt\"))\n",
    "        if os.path.exists(cn_fname) and isfile(cn_fname):\n",
    "            print(f\"Found data pair: {en_fname} and {cn_fname}\")\n",
    "\n",
    "            with open(en_fname, \"r\") as enfin:\n",
    "                en_messages = enfin.read()\n",
    "\n",
    "            with open(cn_fname, \"r\") as cnfin:\n",
    "                cn_messages = cnfin.read()\n",
    "            \n",
    "            en_messages = [part.strip() for part in en_messages.split(\"\\n\\n\") if part.strip() != \"\"]\n",
    "            cn_messages = [part.strip() for part in cn_messages.split(\"\\n\\n\") if part.strip() != \"\"]\n",
    "\n",
    "            if(len(en_messages) != len(cn_messages)):\n",
    "                print(f\"English and Chinese version mismatch. Discarding {en_fname} pair.\")\n",
    "            \n",
    "            # Write document\n",
    "            history = []\n",
    "            for en, cn in zip(en_messages, cn_messages):\n",
    "                all_datapoints.append({\n",
    "                    'system': INSTRUCT_CHUNKED_PROMPT, \n",
    "                    'input': en,\n",
    "                    'output': cn,\n",
    "                    'history': history\n",
    "                })\n",
    "                history.append([en, cn])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "shuffled = random.shuffle(all_datapoints)\n",
    "\n",
    "COUNT = len(all_datapoints)\n",
    "TRAIN = int(0.8 * COUNT)\n",
    "\n",
    "train_dataset = all_datapoints[:TRAIN]\n",
    "evaluation_dataset = all_datapoints[TRAIN:]\n",
    "\n",
    "with open(\"train.json\", \"w\") as train_f:\n",
    "    for d in train_dataset:\n",
    "        json.dump(d, train_f)\n",
    "        train_f.write('\\n')\n",
    "\n",
    "with open(\"evaluation.json\", \"w\") as ev_f:\n",
    "    for d in evaluation_dataset:\n",
    "        json.dump(d, ev_f)\n",
    "        ev_f.write('\\n')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}