challenging-america-word-gap-prediction-kenlm/testing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "21c9b695",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import csv\n",
    "import regex as re\n",
    "import nltk\n",
    "from collections import Counter, defaultdict\n",
    "import string\n",
    "import unicodedata\n",
    "\n",
    "def clean_text(text): \n",
    "    return re.sub(r\"\\p{P}\", \"\", str(text).lower().replace(\"-\\\\n\", \"\").replace(\"\\\\n\", \" \"))\n",
    "\n",
    "def train_model(data, model):\n",
    "    for _, row in data.iterrows():\n",
    "        words = nltk.word_tokenize(clean_text(row[\"final\"]))\n",
    "        for w1, w2 in nltk.bigrams(words, pad_left=True, pad_right=True):\n",
    "            if w1 and w2:\n",
    "                model[w2][w1] += 1\n",
    "    for w1 in model:\n",
    "        total_count = float(sum(model[w1].values()))\n",
    "        for w2 in model[w1]:\n",
    "            model[w2][w1] /= total_count\n",
    "\n",
    "\n",
    "def predict(word, model):\n",
    "    predictions = dict(model[word])\n",
    "    most_common = dict(Counter(predictions).most_common(5))\n",
    "\n",
    "    total_prob = 0.0\n",
    "    str_prediction = \"\"\n",
    "\n",
    "    for word, prob in most_common.items():\n",
    "        total_prob += prob\n",
    "        str_prediction += f\"{word}:{prob} \"\n",
    "\n",
    "    if not total_prob:\n",
    "        return \"the:0.2 be:0.2 to:0.2 of:0.1 and:0.1 a:0.1 :0.1\"\n",
    "\n",
    "    if 1 - total_prob >= 0.01:\n",
    "        str_prediction += f\":{1-total_prob}\"\n",
    "    else:\n",
    "        str_prediction += f\":0.01\"\n",
    "\n",
    "    return str_prediction\n",
    "\n",
    "\n",
    "def predict_data(read_path, save_path, model):\n",
    "    data = pd.read_csv(\n",
    "        read_path, sep=\"\\t\", error_bad_lines=False, header=None, quoting=csv.QUOTE_NONE\n",
    "    )\n",
    "    with open(save_path, \"w\") as file:\n",
    "        for _, row in data.iterrows():\n",
    "            words = nltk.word_tokenize(clean_text(row[7]))\n",
    "            if len(words) < 3:\n",
    "                prediction = \"the:0.2 be:0.2 to:0.2 of:0.1 and:0.1 a:0.1 :0.1\"\n",
    "            else:\n",
    "                prediction = predict(words[-1], model)\n",
    "            file.write(prediction + \"\\n\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e39473e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"in-header.tsv\") as f:\n",
    "    in_cols = f.read().strip().split(\"\\t\")\n",
    "\n",
    "with open(\"out-header.tsv\") as f:\n",
    "    out_cols = f.read().strip().split(\"\\t\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bde510c9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['FileId', 'Year', 'LeftContext', 'RightContext']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "in_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "0e8b31dd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Word']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "out_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7662d802",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv(\n",
    "    \"train/in.tsv.xz\",\n",
    "    sep=\"\\t\",\n",
    "    on_bad_lines='skip',\n",
    "    header=None,\n",
    "    # names=in_cols,\n",
    "    quoting=csv.QUOTE_NONE,\n",
    ")\n",
    "\n",
    "train_labels = pd.read_csv(\n",
    "    \"train/expected.tsv\",\n",
    "    sep=\"\\t\",\n",
    "    on_bad_lines='skip',\n",
    "    header=None,\n",
    "    # names=out_cols,\n",
    "    quoting=csv.QUOTE_NONE,\n",
    ")\n",
    "\n",
    "train_data = data[[7, 6]]\n",
    "train_data = pd.concat([train_data, train_labels], axis=1)\n",
    "\n",
    "train_data[\"final\"] = train_data[7] + train_data[0] + train_data[6]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c3d2cfec",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bd92ba07",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "model = defaultdict(lambda: defaultdict(lambda: 0))\n",
    "\n",
    "train_model(train_data, model)\n",
    "predict_data(\"dev-0/in.tsv.xz\", \"dev-0/out.tsv\", model)\n",
    "predict_data(\"test-A/in.tsv.xz\", \"test-A/out.tsv\", model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ad23240e",
   "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": "3.10.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}