SAN_Figure6_and_7.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Canada's Monetary Policy Report: if text could speak, what would it say?\n",
    "\n",
    "## This notebook creates figures 6 and 7 in the Staff Analytical Note (Classifier performance and Sentiment).\n",
    "\n",
    "## It applies a neural net approach based on https://github.com/fastai/fastai/blob/master/courses/dl2/imdb.ipynb\n",
    "\n",
    "### This workbook should \"just work\". If it does not, or if you find an error, please email the authors.\n",
    "\n",
    "##### André Binette <abinette@bankofcanada.ca>\n",
    "\n",
    "##### Dmitri Tchebotarev\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---------------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set up the environment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import required modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/res/bina/.conda/envs/mpr/lib/python3.6/site-packages/sklearn/ensemble/weight_boosting.py:29: DeprecationWarning: numpy.core.umath_tests is an internal NumPy module and should not be imported. It will be removed in a future NumPy release.\n",
      "  from numpy.core.umath_tests import inner1d\n"
     ]
    }
   ],
   "source": [
    "import sys, os\n",
    "sys.path.insert(0,os.path.realpath(\"./fastai/old\"))\n",
    "\n",
    "from fastai.text import *\n",
    "\n",
    "import warnings\n",
    "\n",
    "from fastai.plots import *\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import functools\n",
    "\n",
    "from scipy.stats import zscore\n",
    "\n",
    "import requests\n",
    "from io import BytesIO\n",
    "from zipfile import ZipFile\n",
    "from bs4 import BeautifulSoup\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as patches\n",
    "import datetime as dt\n",
    "import matplotlib.dates as mdates\n",
    "from datetime import datetime, timedelta"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.backends.cudnn.enabled = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#This notebook uses pathlib for munging paths, which provides some additional functionality over plain strings.\n",
    "PATH=Path('data/boc/')\n",
    "\n",
    "CLAS_PATH=Path('data/boc_clas/')\n",
    "CLAS_PATH.mkdir(exist_ok=True)\n",
    "\n",
    "LM_PATH=Path('data/boc_lm/')\n",
    "LM_PATH.mkdir(exist_ok=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Creating the training and validation set from all MPR sentences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fine tune the LM model we pass through all MPR sentences, regardless of whether they are\n",
    "#labeled or not, we read in all the sentences.\n",
    "\n",
    "CLASSES = ['neg', 'pos', 'unsup', 'meh']\n",
    "\n",
    "def get_texts(path):\n",
    "    texts,labels = [],[]\n",
    "    for idx,label in enumerate(CLASSES):\n",
    "        with open(str(path/(label+\".txt\")), \"tr\") as f:\n",
    "            for line in f:\n",
    "                texts.append(\"xbos xfld 1 \" + line.rstrip())\n",
    "                labels.append(idx)\n",
    "    return np.array(texts),np.array(labels)\n",
    "\n",
    "trn_texts,trn_labels = get_texts(PATH/'train')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14921"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Total number of sentences\n",
    "len(trn_texts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We use a random permutation np array to shuffle the sentences."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(np.prod([ord(c) for c in 'deaa']))\n",
    "trn_idx = np.random.permutation(len(trn_texts))\n",
    "trn_texts = trn_texts[trn_idx]\n",
    "trn_labels = trn_labels[trn_idx]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Then use scikit-learn to split into a training and validation set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_texts,val_texts = sklearn.model_selection.train_test_split(\n",
    "    np.concatenate([trn_texts]), test_size=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(13428, 1493)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(trn_texts), len(val_texts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Format the data and store them as CSV files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "col_names = ['labels','text']\n",
    "\n",
    "df_trn = pd.DataFrame({'text':trn_texts, 'labels':[0]*len(trn_texts)}, columns=col_names)\n",
    "df_val = pd.DataFrame({'text':val_texts, 'labels':[0]*len(val_texts)}, columns=col_names)\n",
    "\n",
    "df_trn.to_csv(LM_PATH/'train.csv', header=False, index=False)\n",
    "df_val.to_csv(LM_PATH/'test.csv', header=False, index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "----------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Language model tokens"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this section, we use the [spacy](http:https://spacy.io) library to tokenize the data. Since spacy does not provide a parallel/multicore version of the tokenizer, the fastai library adds this functionality. This parallel version uses all the cores of your CPUs and runs much faster than the serial version of the spacy tokenizer.\n",
    "\n",
    "Tokenization is the process of splitting the text into separate tokens so that each token can be assigned a unique index. This means we can convert the text into integer indexes which our models can use.\n",
    "\n",
    "We use an appropriate chunksize as the tokenization process is memory intensive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_tokenized_texts(df, n_lbls=1, tokenizer=Tokenizer()):\n",
    "    labels = df.iloc[:,range(n_lbls)].values.astype(np.int64)\n",
    "    texts = df.iloc[:,1].astype(str)\n",
    "    tok = [tokenizer.proc_text(s) for s in texts]\n",
    "    return tok, list(labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_trn = pd.read_csv(LM_PATH/'train.csv', header=None)#, chunksize=chunksize)\n",
    "df_val = pd.read_csv(LM_PATH/'test.csv', header=None)#, chunksize=chunksize)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "tok_trn, trn_labels = get_tokenized_texts(df_trn)\n",
    "tok_val, val_labels = get_tokenized_texts(df_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('the', 25059),\n",
       " (',', 15700),\n",
       " ('in', 15426),\n",
       " ('1', 13844),\n",
       " ('.', 13431),\n",
       " ('xbos', 13428),\n",
       " ('xfld', 13428),\n",
       " ('of', 11814),\n",
       " ('to', 10614),\n",
       " ('and', 8631)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "freq = Counter(p for o in tok_trn for p in o)\n",
    "freq.most_common(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_vocab = 3000\n",
    "min_freq = 2\n",
    "\n",
    "itos = [o for o,c in freq.most_common(max_vocab) if c>min_freq]\n",
    "itos.insert(0, '_pad_')\n",
    "itos.insert(0, '_unk_')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3002"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(itos)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We create a reverse mapping called stoi which is useful to lookup the index of a given token. stoi also has the same number of elements as itos. We use a high performance container called [collections.defaultdict](https://docs.python.org/2/library/collections.html#collections.defaultdict) to store our stoi mapping."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "stoi = collections.defaultdict(lambda:0, {v:k for k,v in enumerate(itos)})\n",
    "vs = len(itos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_lm = np.array([[stoi[o] for o in p] for p in tok_trn])\n",
    "val_lm = np.array([[stoi[o] for o in p] for p in tok_val])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## wikitext103 conversion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are now going to build an english language model. We could start from scratch and try to learn the structure of the english language. But we use a technique called transfer learning to make this process easier. In transfer learning a pre-trained LM that has been trained on a large generic corpus(_like wikipedia articles_) can be used to transfer it's knowledge to a target LM and the weights can be fine-tuned.\n",
    "\n",
    "Our source LM is the wikitext103 LM created by Stephen Merity @ Salesforce research. [Link to dataset](https://www.salesforce.com/products/einstein/ai-research/the-wikitext-dependency-language-modeling-dataset/)\n",
    "The language model for wikitext103 (AWD LSTM) has been pre-trained and the weights can be downloaded here: [Link](http:https://files.fast.ai/models/wt103/). Our target LM is the IMDB LM. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#! wget -nH -r -np -P {PATH} http:https://files.fast.ai/models/wt103/"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The pre-trained LM weights have an embedding size of 400, 1150 hidden units and just 3 layers. We need to match these values  with the target LM so that the weights can be loaded up."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "em_sz,nh,nl = 400,1150,3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "PRE_PATH = PATH/'models'/'wt103'\n",
    "PRE_LM_PATH = PRE_PATH/'fwd_wt103.h5'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "wgts = torch.load(PRE_LM_PATH, map_location=lambda storage, loc: storage)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the mean of the layer0 encoder weights. This can be used to assign weights to unknown tokens when we transfer to target LM."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "enc_wgts = to_np(wgts['0.encoder.weight'])\n",
    "row_m = enc_wgts.mean(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "itos2 = pickle.load((PRE_PATH/'itos_wt103.pkl').open('rb'))\n",
    "stoi2 = collections.defaultdict(lambda:-1, {v:k for k,v in enumerate(itos2)})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before we try to transfer the knowledge from wikitext to the LM, we match up the vocab words and their indexes. \n",
    "We use the defaultdict container once again, to assign mean weights to unknown tokens that do not exist in wikitext103."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_w = np.zeros((vs, em_sz), dtype=np.float32)\n",
    "for i,w in enumerate(itos):\n",
    "    r = stoi2[w]\n",
    "    new_w[i] = enc_wgts[r] if r>=0 else row_m"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We now overwrite the weights into the wgts odict.\n",
    "The decoder module, which we will explore in detail is also loaded with the same weights due to an idea called weight tying."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "wgts['0.encoder.weight'] = T(new_w)\n",
    "wgts['0.encoder_with_dropout.embed.weight'] = T(np.copy(new_w))\n",
    "wgts['1.decoder.weight'] = T(np.copy(new_w))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we have the weights prepared, we are ready to create and start training our new language pytorch model!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Language model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is fairly straightforward to create a new language model using the fastai library. Our model will have a backbone and a custom head. The backbone in our case is the LM pre-trained with wikitext and the custom head is a linear classifier. In this section we will focus on the backbone LM and the next section will talk about the classifier custom head.\n",
    "\n",
    "bptt (*also known traditionally in NLP LM as ngrams*) in fastai LMs is approximated to a std. deviation around 70, by perturbing the sequence length on a per-batch basis. This is akin to shuffling our data in computer vision, only that in NLP we cannot shuffle inputs and we have to maintain statefulness. \n",
    "\n",
    "Since we are predicting words using ngrams, we want our next batch to line up with the end-points of the previous mini-batch's items. batch-size is constant and but the fastai library expands and contracts bptt each mini-batch using a clever stochastic implementation of a batch. (original credits attributed to [Smerity](https://twitter.com/jeremyphoward/status/980227258395770882))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "wd=1e-7\n",
    "bptt=70\n",
    "bs=64\n",
    "opt_fn = partial(optim.Adam, betas=(0.8, 0.99))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The goal of the LM is to learn to predict a word/token given a preceeding set of words(tokens). We take all the sentences in both the training set and validation set and concatenate them to form long strings of tokens. In fastai, we use the `LanguageModelLoader` to create a data loader which makes it easy to create and use bptt sized mini batches. The  `LanguageModelLoader` takes a concatenated string of tokens and returns a loader.\n",
    "\n",
    "We have a special modeldata object class for LMs called `LanguageModelData` to which we can pass the training and validation loaders and get in return the model itself."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_dl = LanguageModelLoader(np.concatenate(trn_lm), bs, bptt)\n",
    "val_dl = LanguageModelLoader(np.concatenate(val_lm), bs, bptt)\n",
    "md = LanguageModelData(PATH, 1, vs, trn_dl, val_dl, bs=bs, bptt=bptt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We setup the dropouts for the model - these values have been chosen after experimentation. If you need to update them for custom LMs, you can change the weighting factor (0.7 here) based on the amount of data you have. For more data, you can reduce dropout factor and for small datasets, you can reduce overfitting by choosing a higher dropout factor. *No other dropout value requires tuning*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "drops = np.array([0.25, 0.1, 0.2, 0.02, 0.15])*1#0.7"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We first tune the last embedding layer so that the missing tokens initialized with mean weights get tuned properly. So we freeze everything except the last layer.\n",
    "\n",
    "We also keep track of the *accuracy* metric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "learner= md.get_model(opt_fn, em_sz, nh, nl, \n",
    "    dropouti=drops[0], dropout=drops[1], wdrop=drops[2], dropoute=drops[3], dropouth=drops[4])\n",
    "\n",
    "learner.metrics = [accuracy]\n",
    "learner.freeze_to(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "learner.model.load_state_dict(wgts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We set learning rates and fit our LM. We first run one epoch to tune the last layer which contains the embedding weights. This should help the missing tokens in the wikitext103 learn better weights."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr=4e-3\n",
    "lrs = lrs = np.array([lr/6,lr/3,lr,lr/2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "510e24ee3ef34489ba6e1d85fa4f4671",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=50, style=ProgressStyle(description_width='initia…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                 \n",
      "    0      4.738204   4.157265   0.321488  \n",
      "    1      4.305157   3.720086   0.349857                 \n",
      "    2      3.884904   3.344446   0.37704                  \n",
      "    3      3.52687    3.060042   0.419873                 \n",
      "    4      3.29097    2.900182   0.439794                 \n",
      "    5      3.124517   2.799553   0.448869                 \n",
      "    6      3.027772   2.730043   0.456418                 \n",
      "    7      2.94754    2.68188    0.461483                 \n",
      "    8      2.879241   2.638991   0.465901                 \n",
      "    9      2.844203   2.61188    0.469015                 \n",
      "    10     2.795701   2.585281   0.472044                 \n",
      "    11     2.766382   2.563747   0.473917                 \n",
      "    12     2.738594   2.551506   0.476207                 \n",
      "    13     2.718729   2.538289   0.476987                 \n",
      "    14     2.699024   2.526093   0.477679                 \n",
      "    15     2.669751   2.512039   0.479321                 \n",
      "    16     2.656446   2.500982   0.481314                 \n",
      "    17     2.650398   2.493937   0.482002                 \n",
      "    18     2.634927   2.488609   0.481955                 \n",
      "    19     2.615983   2.482681   0.482967                 \n",
      "    20     2.600868   2.474046   0.485092                 \n",
      "    21     2.593789   2.472581   0.484782                 \n",
      "    22     2.584435   2.461262   0.486751                 \n",
      "    23     2.582088   2.457217   0.487805                 \n",
      "    24     2.57305    2.459123   0.486588                 \n",
      "    25     2.56954    2.448775   0.488603                 \n",
      "    26     2.565285   2.45016    0.489334                 \n",
      "    27     2.547801   2.445893   0.48941                  \n",
      "    28     2.553057   2.446452   0.489124                 \n",
      "    29     2.551499   2.443151   0.490055                 \n",
      "    30     2.534768   2.428988   0.492128                 \n",
      "    31     2.52566    2.437196   0.491152                 \n",
      "    32     2.531784   2.438042   0.490809                 \n",
      "    33     2.530458   2.437806   0.490263                 \n",
      "    34     2.523447   2.437273   0.490344                 \n",
      "    35     2.53741    2.429826   0.491813                 \n",
      "    36     2.517157   2.428189   0.491663                 \n",
      "    37     2.514964   2.428562   0.491553                 \n",
      "    38     2.512428   2.42796    0.491473                 \n",
      "    39     2.505479   2.424917   0.492119                 \n",
      "    40     2.499404   2.422417   0.492851                 \n",
      "    41     2.495194   2.422473   0.492322                 \n",
      "    42     2.49891    2.422948   0.492315                 \n",
      "    43     2.501581   2.423237   0.492211                 \n",
      "    44     2.487351   2.421868   0.492584                 \n",
      "    45     2.499552   2.415832   0.4931                   \n",
      "    46     2.492726   2.416829   0.493619                 \n",
      "    47     2.496556   2.417392   0.492809                 \n",
      "    48     2.492126   2.42077    0.49229                  \n",
      "    49     2.502481   2.417208   0.493284                 \n",
      "\n",
      "CPU times: user 3d 17h 45min 45s, sys: 26min 28s, total: 3d 18h 12min 14s\n",
      "Wall time: 4h 14min\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[2.4172083536783853, 0.4932840234703488]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "learner.fit(lrs, 1, wds=wd, use_clr=(32,10), cycle_len=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "learner.save('lm1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "learner.save_encoder('lm1_enc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "try: del learner\n",
    "except: pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classifier tokens"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The classifier model is basically a linear layer custom head on top of the LM backbone. Setting up the classifier data is similar to the LM data setup except that we cannot use the unsup sentences this time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "CLASSES = ['neg', 'pos','meh']\n",
    "\n",
    "trn_texts,trn_labels = get_texts(PATH/'train')\n",
    "# val_texts,val_labels = get_texts(PATH/'test')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "900"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(trn_texts)#,len(val_texts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "col_names = ['labels','text']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We use a random permutation np array to shuffle the text reviews."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(np.prod([ord(c) for c in 'deaa']))\n",
    "trn_idx = np.random.permutation(len(trn_texts))\n",
    "# val_idx = np.random.permutation(len(val_texts))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_texts = trn_texts[trn_idx]\n",
    "# val_texts = val_texts[val_idx]\n",
    "\n",
    "trn_labels = trn_labels[trn_idx]\n",
    "# val_labels = val_labels[val_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_trn = pd.DataFrame({'text':trn_texts, 'labels':trn_labels}, columns=col_names)\n",
    "df_val = df_trn.sample(len(df_trn)//10)\n",
    "df_trn = df_trn[~df_trn.index.isin(df_val.index)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "tok_trn, trn_labels = get_tokenized_texts(df_trn)\n",
    "tok_val, val_labels = get_tokenized_texts(df_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_clas = np.array([[stoi[o] for o in p] for p in tok_trn])\n",
    "val_clas = np.array([[stoi[o] for o in p] for p in tok_val])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_labels = np.squeeze(trn_labels)\n",
    "val_labels = np.squeeze(val_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can create our final model, a classifier which is really a custom linear head over our trained backbone. The steps to create the classifier model are similar to the ones for the LM."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "bptt,em_sz,nh,nl = 70,400,1150,3\n",
    "vs = len(itos)\n",
    "opt_fn = partial(optim.Adam, betas=(0.8, 0.99))\n",
    "bs = 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "min_lbl = trn_labels.min()\n",
    "trn_labels -= min_lbl\n",
    "val_labels -= min_lbl\n",
    "c=int(trn_labels.max())+1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the classifier, unlike LM, we need to read a sentence at a time and learn to predict it's sentiment as pos/neg. We do not deal with equal bptt size batches, so we have to pad the sequences to the same length in each batch. To create batches of similar sized, we use a sortish sampler method invented by [@Smerity](https://twitter.com/Smerity) and [@jekbradbury](https://twitter.com/jekbradbury)\n",
    "\n",
    "The sortishSampler cuts down the overall number of padding tokens the classifier ends up seeing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn_ds = TextDataset(trn_clas, trn_labels)\n",
    "val_ds = TextDataset(val_clas, val_labels)\n",
    "trn_samp = SortishSampler(trn_clas, key=lambda x: len(trn_clas[x]), bs=bs//2)\n",
    "val_samp = SortSampler(val_clas, key=lambda x: len(val_clas[x]))\n",
    "trn_dl = DataLoader(trn_ds, bs//2, transpose=True, num_workers=1, pad_idx=1, sampler=trn_samp)\n",
    "val_dl = DataLoader(val_ds, bs, transpose=True, num_workers=1, pad_idx=1, sampler=val_samp)\n",
    "md = ModelData(PATH, trn_dl, val_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "810"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(trn_ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "90"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(val_ds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training the language model and the classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "dps = np.array([0.4, 0.5, 0.05, 0.3, 0.1])*1.4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "dps = np.array([0.4,0.5,0.05,0.3,0.4])*1.35"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "m = get_rnn_classifer(bptt, 20*70, c, vs, emb_sz=em_sz, n_hid=nh, n_layers=nl, pad_token=1,\n",
    "          layers=[em_sz*3, 50, c], drops=[dps[4], 0.1],\n",
    "          dropouti=dps[0], wdrop=dps[1], dropoute=dps[2], dropouth=dps[3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt_fn = partial(optim.Adam, betas=(0.7, 0.99))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = RNN_Learner(md, TextModel(to_gpu(m)), opt_fn=opt_fn)\n",
    "learn.reg_fn = partial(seq2seq_reg, alpha=2, beta=1)\n",
    "learn.clip=25.\n",
    "learn.metrics = [accuracy]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr=3e-3\n",
    "lrm = 2.6\n",
    "lrs = np.array([lr/(lrm**4), lr/(lrm**3), lr/(lrm**2), lr/lrm, lr])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "lrs=np.array([1e-4,1e-4,1e-4,1e-3,1e-2])*2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "wd = 1e-6\n",
    "# wd = 0\n",
    "learn.load_encoder('lm1_enc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.freeze_to(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=UserWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cb02bdb660834b68945274f085bcd27b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=20, style=ProgressStyle(description_width='initia…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                 \n",
      "    0      1.113984   1.040521   0.566667  \n",
      "    1      1.068938   0.89921    0.611111                 \n",
      "    2      1.037036   0.845945   0.566667                 \n",
      "    3      1.015312   0.817725   0.6                      \n",
      "    4      0.989748   0.773899   0.633333                  \n",
      "    5      0.986382   0.762966   0.655556                  \n",
      "    6      0.980703   0.777572   0.688889                  \n",
      "    7      0.979219   0.771307   0.688889                  \n",
      "    8      0.974916   0.790074   0.7                       \n",
      "    9      0.966882   0.75254    0.688889                  \n",
      "    10     0.967451   0.775707   0.666667                  \n",
      "    11     0.968132   0.786648   0.666667                  \n",
      "    12     0.957135   0.756763   0.655556                  \n",
      "    13     0.949537   0.771805   0.644444                  \n",
      "    14     0.922454   0.78357    0.644444                  \n",
      "    15     0.918733   0.756716   0.655556                  \n",
      "    16     0.915544   0.747172   0.644444                  \n",
      "    17     0.914841   0.748273   0.611111                  \n",
      "    18     0.9257     0.755515   0.633333                  \n",
      "    19     0.928016   0.751078   0.622222                  \n",
      "\n",
      "CPU times: user 1h 53min 26s, sys: 1.28 s, total: 1h 53min 27s\n",
      "Wall time: 4min 13s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.7510779950353834, 0.6222222288449605]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "learn.fit(lrs, 1, wds=wd, cycle_len=20, use_clr=(8,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('clas_0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('clas_0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.freeze_to(-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6667267c2c604302be81ed183d025c87",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=20, style=ProgressStyle(description_width='initia…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                  \n",
      "    0      0.920075   0.738577   0.655556  \n",
      "    1      0.926131   0.732942   0.666667                  \n",
      "    2      0.937815   0.727418   0.622222                  \n",
      "    3      0.931443   0.71645    0.655556                  \n",
      "    4      0.931675   0.680632   0.744444                  \n",
      "    5      0.891684   0.648179   0.766667                  \n",
      "    6      0.899113   0.669247   0.744444                  \n",
      "    7      0.913757   0.672778   0.8                       \n",
      "    8      0.886137   0.617714   0.855556                  \n",
      "    9      0.884683   0.598212   0.8                       \n",
      "    10     0.871842   0.611193   0.822222                  \n",
      "    11     0.861563   0.589866   0.822222                  \n",
      "    12     0.845505   0.592709   0.844444                  \n",
      "    13     0.837334   0.588254   0.811111                  \n",
      "    14     0.850592   0.598448   0.733333                  \n",
      "    15     0.847119   0.596153   0.755556                  \n",
      "    16     0.844024   0.573277   0.788889                  \n",
      "    17     0.84166    0.574762   0.777778                  \n",
      "    18     0.836993   0.580573   0.744444                  \n",
      "    19     0.835126   0.599203   0.766667                  \n",
      "\n",
      "CPU times: user 2h 38min 17s, sys: 2.68 s, total: 2h 38min 20s\n",
      "Wall time: 5min 54s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.5992028673489889, 0.766666669315762]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "learn.fit(lrs, 1, wds=wd, cycle_len=20, use_clr=(8,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('clas_1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('clas_1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5f9e4fbade064db39d10477da2195838",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=60, style=ProgressStyle(description_width='initia…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                  \n",
      "    0      0.810904   0.564955   0.8       \n",
      "    1      0.833903   0.525123   0.844444                  \n",
      "    2      0.830276   0.528322   0.855556                  \n",
      "    3      0.813517   0.509737   0.866667                  \n",
      "    4      0.796316   0.46778    0.866667                  \n",
      "    5      0.794056   0.455714   0.866667                  \n",
      "    6      0.780719   0.446851   0.866667                  \n",
      "    7      0.769664   0.420705   0.877778                  \n",
      "    8      0.761432   0.398866   0.877778                  \n",
      "    9      0.734596   0.377794   0.888889                  \n",
      "    10     0.720827   0.340944   0.911111                  \n",
      "    11     0.675594   0.34536    0.888889                  \n",
      "    12     0.686546   0.353657   0.888889                  \n",
      "    13     0.686168   0.382523   0.877778                  \n",
      "    14     0.676448   0.375527   0.877778                  \n",
      "    15     0.647493   0.333193   0.933333                  \n",
      "    16     0.663525   0.325094   0.911111                  \n",
      "    17     0.642945   0.332395   0.911111                  \n",
      "    18     0.63713    0.326396   0.9                       \n",
      "    19     0.625192   0.322097   0.911111                  \n",
      "    20     0.616841   0.318649   0.911111                  \n",
      "    21     0.610617   0.311325   0.922222                  \n",
      "    22     0.589533   0.312072   0.922222                  \n",
      "    23     0.581557   0.30015    0.922222                  \n",
      "    24     0.585995   0.295468   0.922222                  \n",
      "    25     0.577261   0.279271   0.922222                  \n",
      "    26     0.577968   0.276723   0.922222                  \n",
      "    27     0.562697   0.273861   0.922222                  \n",
      "    28     0.552588   0.279413   0.922222                  \n",
      "    29     0.567605   0.276661   0.922222                  \n",
      "    30     0.572208   0.274624   0.911111                  \n",
      "    31     0.566626   0.274439   0.922222                  \n",
      "    32     0.570057   0.293535   0.933333                  \n",
      "    33     0.574891   0.294689   0.9                       \n",
      "    34     0.563017   0.26121    0.933333                  \n",
      "    35     0.550146   0.228923   0.944444                  \n",
      "    36     0.527014   0.24036    0.933333                  \n",
      "    37     0.51636    0.249284   0.933333                  \n",
      "    38     0.523162   0.238933   0.944444                  \n",
      "    39     0.512521   0.259803   0.922222                  \n",
      "    40     0.483988   0.245102   0.922222                  \n",
      "    41     0.472019   0.240221   0.922222                  \n",
      "    42     0.477946   0.24171    0.944444                  \n",
      "    43     0.477769   0.250494   0.933333                  \n",
      "    44     0.459429   0.228319   0.933333                  \n",
      "    45     0.476583   0.263361   0.922222                  \n",
      "    46     0.462663   0.238443   0.922222                  \n",
      "    47     0.455542   0.258749   0.922222                  \n",
      "    48     0.452304   0.258866   0.933333                  \n",
      "    49     0.42681    0.232707   0.933333                  \n",
      "    50     0.425924   0.237544   0.933333                  \n",
      "    51     0.434186   0.255673   0.922222                  \n",
      "    52     0.439159   0.253354   0.922222                  \n",
      "    53     0.426913   0.230224   0.933333                  \n",
      "    54     0.421923   0.230365   0.922222                  \n",
      "    55     0.413447   0.227564   0.922222                  \n",
      "    56     0.418021   0.232833   0.933333                  \n",
      "    57     0.409185   0.236567   0.933333                  \n",
      "    58     0.414863   0.251173   0.944444                  \n",
      "    59     0.422792   0.233276   0.933333                  \n",
      "\n",
      "CPU times: user 14h 28min 36s, sys: 9min 36s, total: 14h 38min 12s\n",
      "Wall time: 38min 21s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.2332757121986813, 0.933333326710595]"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "learn.fit(lrs*1, 2, wds=wd, cycle_len=30, use_clr=(8,8))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('clas_2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('clas_2')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predictions and performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "@functools.lru_cache(maxsize=2**17)\n",
    "def pred(sent, tokenizer=Tokenizer()):\n",
    "    l = [stoi[p] for p in tokenizer.proc_text(sent)]\n",
    "    arr = np.transpose(np.array([l]))    \n",
    "    p = learn.predict_array(arr)[0]\n",
    "    return F.softmax(T(p)).detach().cpu().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In sample\n",
      "Accuracy: 0.9740740740740741 \n",
      "\n",
      "[[263   2   1]\n",
      " [  7 268   1]\n",
      " [  3   7 258]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"In sample\")\n",
    "t = df_trn['labels']\n",
    "p = np.argmax(np.concatenate([pred(txt) for txt in df_trn.text]), axis=1)\n",
    "cm = confusion_matrix(t,p)\n",
    "print(\"Accuracy:\", np.diag(cm).sum()/cm.sum(), \"\\n\")\n",
    "plot_confusion_matrix(cm,['Negative','Neutral','Positive'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In validation\n",
      "Accuracy: 0.9333333333333333 \n",
      "\n",
      "[[32  2  0]\n",
      " [ 0 23  1]\n",
      " [ 1  2 29]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"In validation\")\n",
    "t = df_val['labels']\n",
    "p = np.argmax(np.concatenate([pred(txt) for txt in df_val.text]), axis=1)\n",
    "cm = confusion_matrix(t,p)\n",
    "print(\"Accuracy:\", np.diag(cm).sum()/cm.sum(), \"\\n\")\n",
    "plot_confusion_matrix(cm,['Negative','Neutral','Positive'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 6 - classifier perfomance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "# loading the test set\n",
    "test_set = pd.read_csv('100labeled.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set['pred'] = test_set.text.apply(lambda x:np.argmax(pred(x)))\n",
    "test_set['pred'] = test_set['pred'].map({0:'neg',1:'pos',2:'neutral'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.83 \n",
      "\n",
      "[[34  1  3]\n",
      " [ 3 19  4]\n",
      " [ 5  1 30]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_cm = confusion_matrix(test_set.value,test_set.pred)\n",
    "print(\"Accuracy:\", np.diag(test_cm).sum()/test_cm.sum(), \"\\n\")\n",
    "plot_confusion_matrix(test_cm,['Negative','Neutral','Positive'])\n",
    "\n",
    "plt.savefig('fig6.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Calculating the sentiment for each MPR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "files = os.listdir('mpr_text/')\n",
    "files = [f for f in files if f.endswith('.txt')]\n",
    "files2 = []\n",
    "for file in files:\n",
    "    y,mm,d=map(int,file.split('.')[0].split('-')[1:])\n",
    "    if y<2009 and mm in (1,2,7,8):\n",
    "        continue\n",
    "    if y==2009 and mm==1:\n",
    "        continue\n",
    "    files2.append(file)\n",
    "files=files2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "dates = list(sorted(['-'.join(f.split('.')[0].split('-')[1:]) for f in files]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentiment = pd.DataFrame(index=pd.DatetimeIndex(dates),columns=['Sentiment'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = spacy.load('en_core_web_lg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_sent_mpr(date):\n",
    "    file = 'mpr-'+date+'.txt'\n",
    "    with open('mpr_text/'+file) as ff:\n",
    "        text = ff.read()\n",
    "    sents = list(nlp(text).sents)\n",
    "    sents = [n.text for n in sents]\n",
    "    pos = 0\n",
    "    neg = 0\n",
    "    neutral = 0\n",
    "    for s in sents:\n",
    "        tone = np.argmax(pred(s))\n",
    "        if tone == 0:\n",
    "            neg += 1\n",
    "        elif tone == 1:\n",
    "            pos += 1\n",
    "        elif tone == 2:\n",
    "            neutral += 1\n",
    "    sentim = (pos - neg)/(pos + neg + neutral)\n",
    "    return sentim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentiment.loc[:,'Sentiment'] = list(map(get_sent_mpr,dates))\n",
    "sentiment = sentiment.apply(zscore)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure 7 - Sentiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x648 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Making the chart\n",
    "\n",
    "fig, ((ax1, ax2)) = plt.subplots(2, 1, sharex=True, gridspec_kw = {'height_ratios':[3, 1]})\n",
    "sentiment.plot(ax=ax1,figsize=(16,9),color='k')\n",
    "\n",
    "bottom_tercile = pd.DataFrame(index=sentiment.index, columns=[''],data=np.percentile(sentiment.mean(axis=1),(100/3)*1,interpolation=\"nearest\"))\n",
    "top_tercile = pd.DataFrame(index=sentiment.index, columns=[''],data=np.percentile(sentiment.mean(axis=1),(100/3)*2,interpolation=\"nearest\"))\n",
    "\n",
    "bottom_tercile.plot(figsize=(16,9),alpha=1,ax=ax1,color='gray',linewidth=0.25)\n",
    "top_tercile.plot(figsize=(16,9),alpha=1,ax=ax1,color='gray',linewidth=0.25)\n",
    "\n",
    "\n",
    "ax1.axhline(0,color='k', linewidth=0.025)\n",
    "\n",
    "ax1.axvline(dt.date(2001,2,1),color='lightgray', zorder=20,linewidth=.2)\n",
    "ax1.axvline(dt.date(2008,2,1),color='lightgray', zorder=20,linewidth=.2)\n",
    "ax1.axvline(dt.date(2013,6,3),color='lightgray', zorder=20,linewidth=.2)\n",
    "\n",
    "plt.subplots_adjust(hspace=0)\n",
    "#Import the GDP data from Statcan\n",
    "table_num = \"36100104\"\n",
    "link = f'https://www150.statcan.gc.ca/n1/tbl/csv/{table_num}-eng.zip'\n",
    "r = requests.get(link)\n",
    "zip_file = ZipFile(BytesIO(r.content))\n",
    "GDP_fisher = pd.read_csv(zip_file.open(f'{table_num}.csv'), index_col=10,na_values=['..','...','x','F'])  \n",
    "\n",
    "# Show only the necessary values using V-number\n",
    "GDP_fisher = GDP_fisher.loc['v62305752']\n",
    "Index2 = GDP_fisher.REF_DATE.values\n",
    "GDP_fisher.index = Index2\n",
    "GDP_fisher = GDP_fisher.drop(['REF_DATE', 'GEO', 'DGUID', 'Seasonal adjustment', 'UOM', 'UOM_ID', 'SCALAR_FACTOR', 'SCALAR_ID', 'COORDINATE', 'STATUS', 'SYMBOL', 'TERMINATED', 'DECIMALS'], axis =1)\n",
    "GDP_fisher = GDP_fisher.pivot_table(index = GDP_fisher.index, columns = 'Prices', values = 'VALUE')\n",
    "GDP_fisher.columns = ['real GDP (fisher)']\n",
    "GDP_fisher = GDP_fisher.loc['1994-01':,]\n",
    "\n",
    "# Create the year over year real GDP growth for each quarter \n",
    "GDP_fisher = (GDP_fisher/GDP_fisher.shift(4) - 1)*100\n",
    "GDP_fisher = GDP_fisher.loc['1995-01':,]\n",
    "GDP_fisher.columns = ['Real GDP Growth']\n",
    "\n",
    "# Change to necessary date format for plot \n",
    "GDP_fisher.index = pd.to_datetime(GDP_fisher.index)\n",
    "dates = mdates.date2num(GDP_fisher.index.to_pydatetime())\n",
    "values = GDP_fisher.loc[:,'Real GDP Growth'].values\n",
    "\n",
    "# Plot\n",
    "ymax = 7\n",
    "ymin = -4.5\n",
    "difference = ymax - ymin\n",
    "\n",
    "thiessenstarttime = (dt.date(1995,1,1))\n",
    "thiessenendtime = (dt.date(2001,1,31))\n",
    "thiessenstart = mdates.date2num(thiessenstarttime)\n",
    "thiessenend = mdates.date2num(thiessenendtime)\n",
    "thiessenwidth = thiessenend - thiessenstart \n",
    "\n",
    "dodgestarttime = (dt.date(2001,2,1))\n",
    "dodgeendtime = (dt.date(2008,1,31))\n",
    "dodgestart = mdates.date2num(dodgestarttime)\n",
    "dodgeend = mdates.date2num(dodgeendtime)\n",
    "dodgewidth = dodgeend - dodgestart \n",
    "\n",
    "carneystarttime = (dt.date(2008,2,1))\n",
    "carneyendtime = (dt.date(2013,6,1))\n",
    "carneystart = mdates.date2num(carneystarttime)\n",
    "carneyend = mdates.date2num(carneyendtime)\n",
    "carneywidth = carneyend - carneystart \n",
    "\n",
    "polozstarttime = (dt.date(2013,6,3))\n",
    "polozendtime = (dt.date(2020,6,2))\n",
    "polozstart = mdates.date2num(polozstarttime)\n",
    "polozend = mdates.date2num(polozendtime)\n",
    "polozwidth = polozend - polozstart \n",
    "rectangles = {'Gordon Thiessen' :patches.Rectangle((thiessenstart,.385*difference+ymin),\n",
    "                                                   thiessenwidth,(.25*difference), linewidth = 1,\n",
    "                                                   edgecolor ='w', facecolor='blue', zorder=5, alpha=.2),\n",
    "              'David Dodge': patches.Rectangle((dodgestart,.385*difference+ymin),dodgewidth,\n",
    "                                               (.25*difference), linewidth = 1, edgecolor ='w',\n",
    "                                               facecolor='blue',zorder=5, alpha=.4),\n",
    "              'Mark Carney': patches.Rectangle((carneystart,.385*difference+ymin),carneywidth\n",
    "                                               ,(.25*difference), linewidth = 1, edgecolor ='w',\n",
    "                                               facecolor='navy',zorder=5, alpha=.5),\n",
    "              'Steven S. Poloz': patches.Rectangle((polozstart,.385*difference+ymin),polozwidth,\n",
    "                                                   (.25*difference), linewidth = 1,edgecolor ='w',\n",
    "                                                   facecolor='navy',zorder=5, alpha=.7)}\n",
    "for r in rectangles:\n",
    "    ax2.add_artist(rectangles[r])\n",
    "    rx, ry = rectangles[r].get_xy()\n",
    "    cx = rx + rectangles[r].get_width()/2.0\n",
    "    cy = ry + rectangles[r].get_height()/2.0\n",
    "    ax2.annotate(r, (cx, cy), color='white', weight='bold', \n",
    "                fontsize=15, ha='center', va='center', zorder=20)\n",
    "ax2.set_xlim((dt.date(1995,1,1),dt.date(2020,6,2)))\n",
    "years = pd.to_datetime(range(1995,2020,2), format='%Y')\n",
    "yearnumbers = mdates.date2num(years.to_pydatetime())\n",
    "ax2.set_xticks(yearnumbers)\n",
    "ax2.set_yticks([])\n",
    "starttime = (dt.date(1995,1,1))\n",
    "endtime = (dt.date(2022,1,31))\n",
    "start = mdates.date2num(starttime)\n",
    "end = mdates.date2num(endtime)\n",
    "plt.plot(dates,values, color='w')\n",
    "annotations = sorted([\n",
    "    (dt.date(1997,7,15),'Asian financial crisis',-3.5),\n",
    "    (dt.date(1999,1,1),'Dot-com bubble',4.5),\n",
    "    (dt.date(2001,9,11),'9/11',-3.5),\n",
    "    (dt.date(2002,11,1),'SARS and BSE',4.5),\n",
    "    (dt.date(2008,9,1),'Global financial crisis',-3.5),\n",
    "    (dt.date(2010,9,1),'European debt drisis',4.5),\n",
    "    (dt.date(2012,9,1),'US debt ceiling',-3.5),\n",
    "    (dt.date(2014,9,1),'Oil price shock',4.5),\n",
    "    (dt.date(2016,11,8),'Trade tensions',-3.5)\n",
    "])\n",
    "for q, a in enumerate(annotations):\n",
    "    w = a[2]\n",
    "    dm = mdates.date2num(a[0])\n",
    "    plt.annotate(a[1],xy=(dm,0),xytext=(dm,.5*0 + w),\\\n",
    "            arrowprops=dict(facecolor='black', arrowstyle='-',linewidth=.5),\\\n",
    "            ha='center',va='center')\n",
    "#, fontsize=11, weight='bold'\n",
    "ax2.minorticks_off()\n",
    "ax2.axhline(0,color='k',linewidth=.0025)\n",
    "box1 = ax1.get_position()\n",
    "box2 = ax2.get_position()\n",
    "ax1.set_position([box1.x0, box1.y0 + box1.height * 0.1,\n",
    "                 box1.width, box1.height * 0.9])\n",
    "ax2.set_position([box2.x0, box2.y0+box1.height*0.1,\n",
    "                 box2.width, box2.height])\n",
    "ax1.legend(loc='upper center', bbox_to_anchor=(0.5, -0.25-box2.height),\\\n",
    "          fancybox=True, shadow=False, ncol=4, fontsize = 'medium')\n",
    "ax1.yaxis.tick_right()\n",
    "plt.setp(ax1.get_xticklabels(), fontsize=12)\n",
    "plt.setp(ax1.get_yticklabels(), fontsize=12)\n",
    "ax1.set_ylabel(\"MPR sentiment\", rotation='horizontal', fontsize=12)\n",
    "ax1.yaxis.set_label_coords(0.96,1.02)\n",
    "\n",
    "plt.text(0.125, 0.16, 'Note: SARS is severe acute respiratory syndrome. BSE is bovine spongiform encephalopathy.', transform=plt.gcf().transFigure, fontsize=9.5)\n",
    "plt.text(.775, 0.16, 'Last observation: October 2018 MPR', transform=plt.gcf().transFigure, fontsize=9.5)\n",
    "\n",
    "plt.text(.51, 0.72, 'Top tercile', transform=plt.gcf().transFigure, fontsize=9.5)\n",
    "plt.text(.4, 0.61, 'Bottom tercile', transform=plt.gcf().transFigure, fontsize=9.5)\n",
    "\n",
    "plt.savefig('fig7.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "_draft": {
   "nbviewer_url": "https://gist.github.com/0dd0df21cf404cf2bb51d0148c8b7d8b"
  },
  "gist": {
   "data": {
    "description": "fastai.text imdb example",
    "public": true
   },
   "id": "0dd0df21cf404cf2bb51d0148c8b7d8b"
  },
  "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": "3.6.8"
  },
  "toc": {
   "colors": {
    "hover_highlight": "#DAA520",
    "navigate_num": "#000000",
    "navigate_text": "#333333",
    "running_highlight": "#FF0000",
    "selected_highlight": "#FFD700",
    "sidebar_border": "#EEEEEE",
    "wrapper_background": "#FFFFFF"
   },
   "moveMenuLeft": true,
   "nav_menu": {
    "height": "86px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": true,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false,
   "widenNotebook": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}