notebooks/supervised_models_check.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from collections import defaultdict\n",
    "\n",
    "import fasttext\n",
    "import fasttext.util\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-10_loss-hs_lr-0.2_ngrams-1', '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-10_loss-hs_lr-0.5_ngrams-1', '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-50_loss-hs_lr-0.1_ngrams-1', '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-7_loss-hs_lr-0.1_ngrams-1', '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-7_loss-hs_lr-0.5_ngrams-1', '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-7_loss-hs_lr-0.5_ngrams-2', '../semantic_learning/training_experiments/supervised/fasttext_dim-200_epoch-50_loss-hs_lr-0.1_ngrams-2', '../semantic_learning/training_experiments/supervised/fasttext_dim-500_epoch-50_loss-hs_lr-0.2_ngrams-1']\n",
      "['sunset', 'song', 'music', 'ibiza', 'people', 'school', 'laptop', 'computer', 'sand', 'sky', 'player']\n"
     ]
    }
   ],
   "source": [
    "training_path = '../semantic_learning/training_experiments/supervised/'\n",
    "models = sorted([training_path + file_name for file_name in os.listdir(training_path) if file_name != '.DS_Store'])\n",
    "print(models)\n",
    "words = ['sunset', 'song', 'music', 'ibiza', 'people', 'school', 'laptop', 'computer', 'sand', 'sky', 'player']\n",
    "print(words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Warning : `load_model` does not return WordVectorModel or SupervisedModel any more, but a `FastText` object which is very similar.\n"
     ]
    }
   ],
   "source": [
    "neighbors = defaultdict(lambda: defaultdict(list))\n",
    "for model_name in ['..//semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-7_loss-hs_lr-0.1_ngrams-1']:\n",
    "    model = fasttext.load_model(model_name + '/supervised_model.bin')\n",
    "    # if 'supervised' in model_name:\n",
    "    #     print(model.test('../data/corpus_valid.txt'))\n",
    "    for word in words:\n",
    "        neighbors[model_name][word] = model.get_nearest_neighbors(word)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fasttext_dim-100_epoch-7_loss-hs_lr-0.1_ngrams-1\n",
      "sunset [(0.8578137755393982, 'sunrise'), (0.7734695672988892, 'sunsets'), (0.7710596919059753, 'sunrises'), (0.7059510350227356, 'sun'), (0.6697677969932556, 'horizon'), (0.6607524752616882, 'milky'), (0.6595597267150879, 'dusk'), (0.6586192846298218, 'praising'), (0.6565290093421936, 'moonlight'), (0.6554563045501709, 'skies')]\n",
      "song []\n",
      "music [(0.6502231359481812, 'musicians'), (0.6220690011978149, 'concert')]\n",
      "ibiza []\n",
      "people [(0.7110589742660522, 'grandchildren'), (0.6453851461410522, 'devotees'), (0.6440896391868591, 'youths'), (0.6367863416671753, 'crowds'), (0.6307986378669739, 'refugees'), (0.6246405839920044, 'queue'), (0.6161436438560486, 'percents'), (0.6156197190284729, 'teamwork'), (0.6144832372665405, 'teenagers'), (0.6143847703933716, 'families')]\n",
      "school [(0.6479821801185608, 'schools'), (0.6301422715187073, 'students')]\n",
      "laptop [(0.8151301145553589, 'laptops'), (0.735193133354187, 'computer'), (0.7348428964614868, 'computers'), (0.642896831035614, 'tablet'), (0.6184213161468506, 'technology')]\n",
      "computer [(0.7351933121681213, 'laptop'), (0.698966920375824, 'computers'), (0.6910762786865234, 'screen'), (0.668672502040863, 'technology'), (0.6558332443237305, 'smartphone'), (0.6135727763175964, 'smartphones'), (0.6028887629508972, 'gadgets')]\n",
      "sand [(0.7648510336875916, 'bobs'), (0.732387125492096, 'sands'), (0.6793148517608643, 'airstrip'), (0.645635724067688, 'dune'), (0.6392326951026917, 'beach'), (0.6204648613929749, 'footprints'), (0.6140845417976379, 'dunes'), (0.6138478517532349, 'sandy'), (0.6070677638053894, 'sarong')]\n",
      "sky [(0.770382285118103, 'skies'), (0.7359291911125183, 'moonless'), (0.7278035879135132, 'sunrises'), (0.7209363579750061, 'sunsets'), (0.7042437791824341, 'moonlight'), (0.6871286630630493, 'ufo'), (0.6807305812835693, 'milky'), (0.6796682476997375, 'thunderstorm'), (0.6762917637825012, 'cumulus'), (0.6753883957862854, 'sunrise')]\n",
      "player [(0.6686747074127197, 'footballer'), (0.6166597604751587, 'winger')]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for model_name in neighbors:\n",
    "    print(model_name.split('/')[-1])\n",
    "    for word in words:\n",
    "        print(word, [x for x in neighbors[model_name][word] if x[0] >= 0.6])\n",
    "    print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Warning : `load_model` does not return WordVectorModel or SupervisedModel any more, but a `FastText` object which is very similar.\n"
     ]
    }
   ],
   "source": [
    "selected_model = '../semantic_learning/training_experiments/supervised/fasttext_dim-100_epoch-7_loss-hs_lr-0.1_ngrams-1/supervised_model.bin'\n",
    "reduced_model = fasttext.util.reduce_model(fasttext.load_model(selected_model), 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Warning : `load_model` does not return WordVectorModel or SupervisedModel any more, but a `FastText` object which is very similar.\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1152x1152 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = fasttext.load_model(selected_model)\n",
    "plt.figure(figsize=(16, 16))\n",
    "for word in words:\n",
    "    plt.scatter(*reduced_model[word].tolist())\n",
    "    plt.annotate(word, reduced_model[word].tolist())\n",
    "    for confidence, neighbor in model.get_nearest_neighbors(word):\n",
    "        if confidence >= 0.7:\n",
    "            plt.scatter(*reduced_model[neighbor].tolist())\n",
    "            plt.annotate(neighbor, reduced_model[neighbor].tolist())\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(0.3755949139595032, 'razor'),\n",
       " (0.35709744691848755, 'blustery'),\n",
       " (0.3558627963066101, 'festivalcomedy'),\n",
       " (0.35475972294807434, 'loom'),\n",
       " (0.3521975576877594, 'bagpipes'),\n",
       " (0.33974704146385193, 'jetting'),\n",
       " (0.33838289976119995, 'publishing'),\n",
       " (0.33816710114479065, 'sandwiched'),\n",
       " (0.3307796120643616, 'judicial'),\n",
       " (0.33046358823776245, 'ultimate')]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.get_nearest_neighbors('</s>', k=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.38 ms ± 4.15 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "model.get_nearest_neighbors(word, k=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "daa06ee645250b8a2dc707d71e9800f0b1cc0a90be827c1c2df03a597f042bff"
  },
  "kernelspec": {
   "display_name": "Python 3.8.12 64-bit ('typing-assistant': conda)",
   "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": "3.8.11"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}