validation_interactive.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "### Interactive medword testing tool ###\n",
    "\n",
    "import medword_pipeline\n",
    "import embedding_fasttext\n",
    "import embedding_word2vec\n",
    "import shared.load_config \n",
    "import json\n",
    "import pylab\n",
    "import os\n",
    "import model_validation as mv\n",
    "import preprocess as pp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "### define filepaths and names ###\n",
    "#\n",
    "# Note: the embedding model must fit the embedding_method in \n",
    "#       the config file (\"word2vec\" or \"fasttext\")\n",
    "# \n",
    "\n",
    "# where to get the configuration file corresponding to the embedding model\n",
    "config_data_dir = 'data/embeddings/emb4/'\n",
    "config_filename = 'emb_model_w2v_4_configuration.json'\n",
    "\n",
    "\n",
    "# where to get the trained embedding model\n",
    "emb_model_dir = 'data/embeddings/emb4/'\n",
    "emb_model_filename = 'emb_model_w2v_4.bin'\n",
    "\n",
    "\n",
    "# Load configuration (don't change)\n",
    "config_src = os.path.join(config_data_dir, config_filename)\n",
    "config = shared.load_config.Configuration(config_src)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model loaded.\n"
     ]
    }
   ],
   "source": [
    "### initialize embedding object (don't change)\n",
    "if config.config['embedding_method'] == 'fasttext':\n",
    "    embedding = embedding_fasttext.EmbeddingFasttext(config)\n",
    "\n",
    "elif config.config['embedding_method'] == 'word2vec':\n",
    "    embedding = embedding_word2vec.EmbeddingWord2vec(config)\n",
    "\n",
    "else:\n",
    "    print('embedding_algorithm (in config) must be \"fasttext\" or \"word2vec\"')\n",
    "    raise AttributeError\n",
    "\n",
    "    \n",
    "### Load model (don't change)\n",
    "embedding.load_model(emb_model_dir, emb_model_filename)\n",
    "print(\"Model loaded.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "vocab size: 584831\n",
      "vector dim: 150\n"
     ]
    }
   ],
   "source": [
    "### Get model information ###\n",
    "print(\"vocab size:\", len(embedding.get_vocab()))\n",
    "print(\"vector dim:\", embedding.vec_dim())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "word is in model vocab:  True\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('hausarzt', 0.7083759719272735),\n",
       " ('einweisende', 0.7014030413022272),\n",
       " ('behandelnde', 0.6947822577577334),\n",
       " ('darzt-rezept', 0.6834899853247314),\n",
       " ('gkv-versicherter', 0.6832060187451043),\n",
       " ('unterschreibt', 0.6763688425610666),\n",
       " ('heilhilfsberufe', 0.6760111969065752),\n",
       " ('eilfall', 0.6757200643686898),\n",
       " ('unterschreibe', 0.6755855001100888),\n",
       " ('konsiliargespräch', 0.6743831698015321)]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "### Play with model ###\n",
    "\n",
    "# define a word from which you like to see it's n-nearest neighbours \n",
    "# in the embedding space\n",
    "word = 'arzt'\n",
    "n_nearest_neighbours = 10\n",
    "\n",
    "# check if word is in model vocab\n",
    "print(\"word is in model vocab: \", word in embedding.get_vocab())\n",
    "\n",
    "# create list of neighbours\n",
    "embedding.most_similar_n(word, n_nearest_neighbours)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('jeanine', 0.26620751310760143),\n",
       " ('landsberger', 0.2660433696237113),\n",
       " ('80639', 0.2656330931007473),\n",
       " ('dipl-soz-päd', 0.2637328793164403),\n",
       " ('warmuth', 0.2632840142348419),\n",
       " ('10585', 0.26105168439033044),\n",
       " ('48161', 0.26072782732789107),\n",
       " ('80339', 0.2598002194313916),\n",
       " ('inka', 0.25894442852951605),\n",
       " ('klinikstrasse', 0.25826217955434494)]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "### Analogy\n",
    "\n",
    "# embedding.analogy([positives], [negatives], topn)\n",
    "embedding.analogy(['koenig', 'frau'], ['mann'], 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.08220343  0.08810184 -0.05436552 -0.1531591   0.02436036 -0.13343668\n",
      "  0.06235622  0.09080388  0.05949959  0.00184021]\n"
     ]
    }
   ],
   "source": [
    "# Get a vector of a word\n",
    "vec = embedding.word_vec('das')\n",
    "\n",
    "print(vec[:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start validation. Loading model. \n",
      "\n",
      "Validating 'doesntfit' with file data/validation_data/german_doesntfit1.txt\n",
      "\n",
      "*** Doesn't fit ***\n",
      "Doesn't fit correct:  76.8% (76/99)\n",
      "Doesn't fit coverage: 90.0% (99/110) \n",
      "\n",
      "\n",
      "*** Human-Similarity ***\n",
      "Number of instances: 636, skipped: 143\n",
      "random error: 0.29851118826, mean error: 0.244944473289\n"
     ]
    }
   ],
   "source": [
    "### Validation of the embedding model ###\n",
    "\n",
    "# where to get the validation data\n",
    "config.config['val_data_dir'] = 'data/validation_data/'\n",
    "\n",
    "# Choose DOESNTFIT test file from your validation folder\n",
    "# note: - the doesntfit file has a 'question' on each line\n",
    "#       - a 'question' is of the format \n",
    "#       \n",
    "#         word_1 word_2 word_3 ... word_N word_not_fitting\n",
    "#        \n",
    "#         -> word_1 to word_N belong to a category but word_not_fitting doesn't\n",
    "# \n",
    "#         Example question:\n",
    "# \n",
    "#         Auto Motorrad Fahrrad Fußgänger\n",
    "#\n",
    "#\n",
    "config.config['doesntfit_file'] = 'german_doesntfit1.txt'\n",
    "\n",
    "\n",
    "# Choose SYNONYM test file from your validation folder\n",
    "# note: - the synonym file has a 'question' on each line\n",
    "#       - a 'question' is of the format \n",
    "#       \n",
    "#         word_1 word_2 \n",
    "#        \n",
    "#         -> word_1 and word_N are synonyms\n",
    "# \n",
    "#         Example question:\n",
    "# \n",
    "#         Blutgerinnsel Thrombus\n",
    "#\n",
    "# \n",
    "config.config['synonyms_file'] = 'german_synonyms3.txt'\n",
    "\n",
    "\n",
    "mv.validate_model(embedding, emb_model_dir, emb_model_filename)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAm8AAAHVCAYAAACnjxwGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8f018e2908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "### Visualization ###\n",
    "# From a list of words, plot all words and it's neighbours \n",
    "\n",
    "# define the words you would like to visualize\n",
    "word_list = ['arzt', 'sonne', 'der']\n",
    "\n",
    "# define the number of closest neighbors to display per word\n",
    "n_nearest_neighbours = 4\n",
    "\n",
    "# change figure size if desired\n",
    "width = 10.0\n",
    "height = 8.0\n",
    "pylab.rcParams['figure.figsize'] = (width, height)\n",
    "\n",
    "# plotting\n",
    "mv.visualize_words(embedding, word_list, n_nearest_neighbours)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['asdf', 'asdf', 'abk', 'hey', 'asdf', 'ein', 'arzt', 'bzw', 'eine', 'ärztin', 'ist', 'ein', 'medizinisch', 'ausgebildeter', 'heilkundiger', 'beschäftigt', 'sich', 'mit', 'der', 'vorbeugung', 'prävention', 'erkennung', 'diagnose', 'behandlung', 'therapie', 'und', 'nachsorge', 'von', 'krankheiten', 'und', 'leiden', 'bzw', 'gesundheitlichen', 'beeinträchtigungen', 'und', 'stellt', 'sich', 'damit', 'in', 'den', 'dienst', 'der', 'gesundheit', 'stae-asdf']\n"
     ]
    }
   ],
   "source": [
    "### test your used tokenizer ###\n",
    "\n",
    "tokenizer = pp.get_tokenizer(config)\n",
    "example = \"\"\"!! „asdf“ 'asdf' abk. hey \"asdf\" ??Ein Arzt bzw. eine Ärztin ist ein medizinisch\\n \\n \n",
    "             ausgebildeter Heilkundiger, beschäftigt sich mit der\n",
    "             Vorbeugung (Prävention), [Erkennung] (Diagnose),\n",
    "             Behandlung (Therapie) [...] und Nachsorge von Krankheiten\n",
    "             und Leiden bzw. gesundheitlichen Beeinträchtigungen,\n",
    "             und stellt sich\">\n",
    "             damit in den Dienst der Gesundheit.\n",
    "             \n",
    "             stae-asdf\"\"\"\n",
    "             \n",
    "tk_example = tokenizer.tokenize(str(example))\n",
    "\n",
    "print(tk_example)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import importlib\n",
    "importlib.reload(mv)\n",
    "file_src = 'data/validation_data/human_similarity.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda env:tensorflow]",
   "language": "python",
   "name": "conda-env-tensorflow-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}