medword_pipeline_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 embedding_word2vec_composite\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\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "### define a new configuration\n",
    "config_dict = {\n",
    "    \n",
    "    # changes base data directory for developing on smaller data.\n",
    "    # possible values: \"develop\" and \"normal\"\n",
    "    \"running_mode\": \"normal\",\n",
    "    \n",
    "    # if you want to compute new training data from raw-data\n",
    "    \"compute_new_data\": False,\n",
    "    # note: if you don't compute new training data, you must provide them\n",
    "    #       at train_data_src for training a new model\n",
    "    \n",
    "    # if you want to train a new model on training data\n",
    "    \"train_new_model\": False,\n",
    "    # note: if you don't train a new model, you must provide it\n",
    "    #       at emb_model_src for the validation step\n",
    "    \n",
    "    # if you want to run the validation\n",
    "    \"run_validation\": False,\n",
    "    \n",
    "    # chose embedding method. possible values: \"word2vec\" and \"fasttext\"\n",
    "    \"embedding_method\": \"word2vec-composite\",\n",
    "    \n",
    "    # chose embedding algorithm. possible values: \"skipgram\" and \"cbow\"\n",
    "    \"embedding_algorithm\": \"skipgram\",\n",
    "    \n",
    "    # chose filenames for train data and embedding model\n",
    "    \"train_data_filename\": \"train.txt\",\n",
    "    \"embedding_model_filename\": \"emb_model_ft.bin\",\n",
    "    \n",
    "    # base data dir when running in \"normal\" mode\n",
    "    \"base_data_dir\": \"data/\",\n",
    "    \n",
    "    # base data dir when running in \"develop\" mode\n",
    "    \"develop_base_data_dir\": \"dev_data/\",\n",
    "    \n",
    "    # embedding model settings\n",
    "    \"embedding_vector_dim\": 150,\n",
    "    \"min_token_appearance\": 5,\n",
    "    \n",
    "    # chose tokenizer. possible values: \"nst\" and \"sgt\"\n",
    "    ## NonStemmingTokenizer: 'nst'\n",
    "    # - no stemming, only remove punctuation marks\n",
    "    # - lowercase letters\n",
    "\n",
    "    ## SimpleGermanTokenizer: 'sgt'\n",
    "    # - remove punctuation marks\n",
    "    # - stemming\n",
    "    # - lowercase letters\n",
    "    \"tokenizer\": \"nst\",\n",
    "    \n",
    "    # validation data settings\n",
    "    \"val_data_dir\": \"data/validation_data/\",\n",
    "    \"doesntfit_file\": \"german_doesntfit1.txt\",\n",
    "    \"synonyms_file\": \"german_synonyms_phrases.txt\",\n",
    "    \n",
    "    # where to store the configuration file\n",
    "    \"config_path\": \"configuration-ipython.json\"\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'json' is not defined",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-3-80b95006e4d2>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mopen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig_src\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'w+'\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m     \u001b[0mjson\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdump\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mindent\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m4\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;31m# load config object based on config file (don't change)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'json' is not defined"
     ],
     "output_type": "error"
    }
   ],
   "source": [
    "### save config file (don't change)\n",
    "config_src = config_dict[\"config_path\"]\n",
    "\n",
    "with open(config_src, 'w+') as f:\n",
    "    json.dump(config_dict, f, indent=4)\n",
    "    \n",
    "# load config object based on config file (don't change)\n",
    "config = shared.load_config.Configuration(config_src)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "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",
    "elif config.config['embedding_method'] == 'word2vec-composite':\n",
    "    embedding = embedding_word2vec_composite.EmbeddingWord2vecComposite(config)\n",
    "    \n",
    "\n",
    "else:\n",
    "    print('embedding_algorithm (in config) must be \"fasttext\" or \"word2vec\"')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to /home/tim/nltk_data...\n",
      "[nltk_data]   Package punkt is already up-to-date!\n",
      "Running in NORMAL mode.\n"
     ]
    }
   ],
   "source": [
    "### run pipeline for that embedding (create train data, train model, validate model)\n",
    "medword_pipeline.run_pipeline(embedding)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train or load a model.\n"
     ]
    }
   ],
   "source": [
    "### Play with embedding model ###\n",
    "#\n",
    "# The following parts are intended to get familiar with the \n",
    "# model and the data\n",
    "\n",
    "# check if model is instantiatet\n",
    "if embedding._model is None: \n",
    "    print(\"Train or load a model.\")\n",
    "else: \n",
    "    print(\"Model is present.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model loaded.\n"
     ]
    }
   ],
   "source": [
    "# maybe load model (if not created in training or loaded in validation of pipeline)\n",
    "# Note: The model must fit the embedding method (fasttext or word2vec)!\n",
    "emb_model_dir = 'data/embeddings/'\n",
    "emb_model_filename = 'emb_model_ft.bin'\n",
    "\n",
    "embedding.load_model(emb_model_dir, emb_model_filename)\n",
    "print(\"Model loaded.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "vocab size: 361798\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": 21,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "word is in model vocab:  False\n",
      "cet\n",
      "eco\n",
      "sandoz\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[('langt', 0.8550068976483358),\n",
       " ('roetzer', 0.8442715031884811),\n",
       " ('droesser', 0.8438368622149449),\n",
       " ('schonfrist', 0.8408969563123138),\n",
       " ('film101', 0.8362701290420604),\n",
       " ('zeitmagazin', 0.8358793212009645),\n",
       " ('19972006', 0.8352904899210074),\n",
       " ('weltprint', 0.8341318972235378),\n",
       " ('zeitungsleser', 0.8325057911194269),\n",
       " ('rauterberg', 0.8301941100222336)]"
      ]
     },
     "execution_count": 21,
     "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 = 'ceteco sandoz'\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)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('bartholinitis', 0.5540084583295412),\n",
       " ('pelveoperitonitis', 0.5516970501114253),\n",
       " ('n739', 0.5441945208683527),\n",
       " ('eileiter', 0.5397869951330914),\n",
       " ('adnexitis', 0.5350158147085161),\n",
       " ('beckenadhaesion', 0.53321237220386),\n",
       " ('genitalprolaps', 0.5329878388691829),\n",
       " ('fournier-gangraen', 0.5301762943086812),\n",
       " ('endometritis', 0.5289991945594156),\n",
       " ('epididymitis', 0.527432929956771)]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "### Analogy\n",
    "\n",
    "# embedding.analogy([positives], [negatives], topn)\n",
    "embedding.analogy(['entzuendung', 'frau'], [], 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.05576208,  0.07286525, -0.11488426,  0.06763246,  0.01310543,\n",
       "       -0.00972985, -0.01771902, -0.12122954,  0.20332509,  0.02934829])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Get a vector of a word\n",
    "import numpy as np\n",
    "vec = embedding.word_vec('blinddarmentzuendung')\n",
    "vec[:10]\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqYAAAHVCAYAAAAjJTltAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDKaSJEnKBYOpJEmScuH/B54dDdQzITpyAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0d1335b198>"
      ]
     },
     "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 = ['r651', 'sirs mit organkomplikation', 'r650']\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": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'config' is not defined",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-b107d4b97a0b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;31m# - no stemming, only remove punctuation marks\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;31m# - lowercase letters\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'tokenizer'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m'nst'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;31m## SimpleGermanTokenizer: 'sgt'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'config' is not defined"
     ],
     "output_type": "error"
    }
   ],
   "source": [
    "### test a tokenizer for the prepocessing and validation ###\n",
    "\n",
    "## NonStemmingTokenizer: 'nst'\n",
    "# - no stemming, only remove punctuation marks\n",
    "# - lowercase letters\n",
    "config.config['tokenizer'] = 'nst'\n",
    "\n",
    "## SimpleGermanTokenizer: 'sgt'\n",
    "# - remove punctuation marks\n",
    "# - stemming\n",
    "# - lowercase letters\n",
    "# config.config['tokenizer'] = 'sgt'\n",
    "\n",
    "# test the chosen tokenizer\n",
    "tokenizer = pp.get_tokenizer(config)\n",
    "example = \"sternotomie 35h111\"\n",
    "             \n",
    "tk_example = tokenizer.tokenize(str(example))\n",
    "\n",
    "print(tk_example)"
   ]
  },
  {
   "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
}