experiments/K_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import os\n",
    "import random\n",
    "import numpy as np\n",
    "import gzip\n",
    "import copy\n",
    "from sklearn.manifold import MDS\n",
    "from sklearn import preprocessing\n",
    "from sklearn import decomposition\n",
    "from nltk.stem.snowball import SnowballStemmer\n",
    "stemmer = SnowballStemmer(\"english\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0, os.path.abspath('..'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from klearn import IT\n",
    "from klearn import utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import copy\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as mpatches\n",
    "from matplotlib.lines import Line2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.font_manager as font_manager\n",
    "from matplotlib.lines import Line2D\n",
    "import matplotlib.ticker as mtick\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import seaborn as sns\n",
    "import os\n",
    "fontpath = os.path.expanduser('~/Downloads/LinLibertine_DRah.ttf')\n",
    "prop = font_manager.FontProperties(fname=fontpath)\n",
    "\n",
    "params = {\n",
    "    \"axes.titlesize\" : 16,\n",
    "    'axes.labelsize': 14,\n",
    "    'font.size': 14,\n",
    "    'legend.fontsize': 14,\n",
    "    'xtick.labelsize': 14,\n",
    "    'ytick.labelsize': 12,\n",
    "    'font.family': prop.get_name(),\n",
    "    'text.usetex': True\n",
    "}\n",
    "\n",
    "mpl.rcParams.update(params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "INPUT_FOLDER = '/Users/peyrardm/Documents/NLP/KLearn-preprocess/data_converted'\n",
    "# INPUT_FOLDER = 'data'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset_08 = utils.load_data(os.path.join(INPUT_FOLDER, 'TAC2008_1.pkl'))\n",
    "dataset_09 = utils.load_data(os.path.join(INPUT_FOLDER, 'TAC2009_1.pkl'))\n",
    "dataset = copy.copy(dataset_08)\n",
    "dataset.update(dataset_09)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_json(filepath):\n",
    "    with open(filepath, 'r') as f:\n",
    "        return json.loads(f.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_corr(annotators_K):\n",
    "    keys = annotators_K.keys()\n",
    "    idx = range(len(keys))\n",
    "    keys_idx = dict(zip(keys, idx))\n",
    "    \n",
    "    sizes = []\n",
    "    annots = list(annotators_K.items())\n",
    "    corr = np.ones((len(keys), len(keys)))\n",
    "    for a_i, v_i in annots:\n",
    "        sizes.append(np.mean(utils.evaluate_K(dataset, v_i['K'], 1., 1.)))\n",
    "        for a_j, v_j in annots:\n",
    "            x_i = keys_idx[a_i]\n",
    "            x_j = keys_idx[a_j]\n",
    "            corr[x_i][x_j] = IT.JS(v_i['K'], v_j['K'])\n",
    "    return corr, keys_idx, np.array(sizes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def map_to_dataset(name):\n",
    "    if '8' in name:\n",
    "        return 'TAC-2008'\n",
    "    return 'TAC-2009'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../outputs/assessors_Ks.json'\n",
    "with open(filename, 'r') as f:\n",
    "    annotators_K = json.loads(f.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "corr_ann, keys_idx_ann, sizes_ann = build_corr(annotators_K)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = corr_ann \n",
    "embedding = MDS(n_components=2, dissimilarity='precomputed')\n",
    "X_transformed = embedding.fit_transform(X)\n",
    "    \n",
    "annotators = [f for f, _ in sorted(keys_idx_ann.items(), key=lambda t: t[1])]\n",
    "annotators_dataset = [map_to_dataset(f) for f in annotators]\n",
    "x = X_transformed[:,0]\n",
    "y = X_transformed[:,1]\n",
    "\n",
    "data_annotator_ann = pd.DataFrame.from_dict({'X': x, 'Y': y, 'generalizability': sizes_ann, 'annotators':annotators, 'annotators_dataset': annotators_dataset})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = '../outputs/domain_Ks.json'\n",
    "with open(filename, 'r') as f:\n",
    "    domains_K = json.loads(f.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "news_keys = {'LiveBlogs', 'TAC2008', 'TAC2009', 'NYT', 'CNN-DM', 'X-Sum'}\n",
    "News_subsets = dict([(key, K) for key, K in domains_K.items() if key in news_keys])\n",
    "Non_News_subsets = dict([(key, K) for key, K in domains_K.items() if key not in news_keys])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def map_to_domain(name):\n",
    "    if name in Non_News_subsets:\n",
    "        return 'Non-News'\n",
    "    return 'News'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "corr_dom, keys_idx_dom, sizes_dom = build_corr(domains_K)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = corr_dom\n",
    "embedding = MDS(n_components=2, dissimilarity='precomputed')\n",
    "X_transformed = embedding.fit_transform(X)\n",
    "\n",
    "domain = [f for f, _ in sorted(keys_idx_dom.items(), key=lambda t: t[1])]\n",
    "domain_dataset = [map_to_domain(f) for f in domain]\n",
    "annotators = [f for f, _ in sorted(keys_idx_dom.items(), key=lambda t: t[1])]\n",
    "annotators_dataset = [map_to_domain(f) for f in annotators]\n",
    "x = X_transformed[:,0]\n",
    "y = X_transformed[:,1]\n",
    "    \n",
    "data_domain_MDS = pd.DataFrame.from_dict({'X': x, 'Y': y, 'generalizability': sizes_dom, 'domains':domain, 'domains_dataset': domain_dataset})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set_context('paper')\n",
    "fig, axes = plt.subplots(1, 2, figsize=(10, 5))\n",
    "\n",
    "ax = axes[0]\n",
    "colors = ['tab:blue', 'tab:red', 'tab:grey']\n",
    "sns.scatterplot(x='X', y='Y', size='generalizability', color=colors[1], ax=ax,\n",
    "                        data=data_annotator_ann, sizes=(300,2000), palette=colors, legend=False, alpha=0.8)\n",
    "\n",
    "# for ann, x, y in zip(annotators_K.keys(), data_annotator_ann['X'].to_list(), data_annotator_ann['Y'].to_list()):\n",
    "#     ax.text(x-0.025, y-0.01, ann, color = 'white', fontsize=15)\n",
    "    \n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "ax.spines['left'].set_visible(False)\n",
    "ax.spines['bottom'].set_visible(False)\n",
    "ax.set_title('a) Annotators MDS', y=-0.1, fontsize=21)\n",
    "\n",
    "# ax.grid()\n",
    "ax.set_xticklabels([])\n",
    "ax.set_yticklabels([])\n",
    "ax.set_ylabel('')\n",
    "ax.set_xlabel('')\n",
    "ax.set_xlim([-0.6, 0.6])\n",
    "ax.set_ylim([-0.6, 0.6])\n",
    "for tic in ax.xaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "    \n",
    "for tic in ax.yaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "\n",
    "ax = axes[1]\n",
    "\n",
    "sns.scatterplot(x='X', y='Y', size='generalizability', hue='domains_dataset', ax=ax,\n",
    "                        data=data_domain_MDS, sizes=(200,2000), palette=[colors[0], colors[2]], legend=False, alpha=0.8)\n",
    "\n",
    "for dom, x, y in zip(domains_K.keys(), data_domain_MDS['X'].to_list(), data_domain_MDS['Y'].to_list()):\n",
    "    if dom == 'TAC2008':\n",
    "        dom = 'TAC08'\n",
    "    elif dom == 'TAC2009':\n",
    "        dom = 'TAC09'\n",
    "    elif dom == 'CNN-DM':\n",
    "        dom = 'CNN'\n",
    "    ax.text(x-0.02, y-0.01, dom, color = 'black', fontsize=17)\n",
    "    \n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "ax.spines['left'].set_visible(False)\n",
    "ax.spines['bottom'].set_visible(False)\n",
    "ax.set_title('b) Domains MDS', y=-0.1, fontsize=21)\n",
    "\n",
    "# ax.grid()\n",
    "ax.set_xticklabels([])\n",
    "ax.set_yticklabels([])\n",
    "ax.set_ylabel('')\n",
    "ax.set_xlabel('')\n",
    "\n",
    "for tic in ax.xaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "    \n",
    "for tic in ax.yaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "    \n",
    "#g.get_legend().set_visible(False)\n",
    "legend_elem = [\n",
    "                mpatches.Patch(facecolor=colors[1], label='Annotators', alpha=0.8),\n",
    "                mpatches.Patch(facecolor=colors[0], label='Non-News', alpha=0.8),\n",
    "                mpatches.Patch(facecolor=colors[2], label='News', alpha=0.8)]\n",
    "                   \n",
    "fig.legend(handles=legend_elem, ncol=3, loc='upper center', frameon=False, fontsize=21,  bbox_to_anchor=(0.5, 1.08))\n",
    "ax.set_ylabel('')\n",
    "ax.set_xlabel('')\n",
    "fig.tight_layout(pad=1.05)\n",
    "fig.savefig(\"../outputs/mds.pdf\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "TAC08_ann = [ann for ann in annotators_K.keys() if '8' in ann]\n",
    "TAC09_ann = [ann for ann in annotators_K.keys() if '9' in ann]\n",
    "\n",
    "x,y = [], []\n",
    "nb_samples = 3\n",
    "for nb_avg in range(1, int(len(annotators_K) /2.) , 3):\n",
    "    for _ in range(nb_samples):\n",
    "        ## Select from TAC-08 and evaluate on TAC-09\n",
    "        selected_keys = random.sample(TAC08_ann, nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset_09, average_K, 1, 1)))\n",
    "        \n",
    "        ## Select from TAC-09 and evaluate on TAC-08\n",
    "        selected_keys = random.sample(TAC09_ann, nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset_08, average_K, 1, 1)))\n",
    "        \n",
    "averaging_generalizability_ann_test = pd.DataFrame.from_dict({'# annotators averaged': x, 'Correlation with humans': y})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "x,y = [], []\n",
    "nb_samples = 3\n",
    "for nb_avg in range(1, len(annotators_K), 3):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(annotators_K.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset, average_K, 1, 1)))\n",
    "\n",
    "full_avg = IT.avg([v['K'] for key, v in annotators_K.items()])\n",
    "x.append(len(annotators_K))\n",
    "y.append(np.mean(utils.evaluate_K(dataset, full_avg, 1, 1)))\n",
    "#x.append(len(annotators_K))\n",
    "#y.append(utils.evaluate_K(dataset, full_avg, 1, 1))\n",
    "averaging_generalizability_ann = pd.DataFrame.from_dict({'# annotators averaged': x, 'Correlation with humans': y})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "x,y = [], []\n",
    "nb_samples = 6\n",
    "for nb_avg in range(1, len(Non_News_subsets), 2):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(Non_News_subsets.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in Non_News_subsets.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset, average_K, 1, 1)))\n",
    "\n",
    "#full_avg = IT.avg([K for key, K in Non_News_subsets.items()])\n",
    "#x.append(len(News_subsets))\n",
    "#y.append(pipeline.evaluate_generic_K(dataset, full_avg, 1, 1))\n",
    "averaging_generalizability_non_news = pd.DataFrame.from_dict({'# domains averaged': x, 'Correlation with humans': y})\n",
    "\n",
    "x,y = [], []\n",
    "nb_samples = 6\n",
    "for nb_avg in range(1, len(domains_K), 3):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(domains_K.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in domains_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset, average_K, 1, 1)))\n",
    "\n",
    "#full_avg = IT.avg([K for key, K in Non_News_subsets.items()])\n",
    "#x.append(len(News_subsets))\n",
    "#y.append(pipeline.evaluate_generic_K(dataset, full_avg, 1, 1))\n",
    "averaging_generalizability_dom = pd.DataFrame.from_dict({'# domains averaged': x, 'Correlation with humans': y})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('/Users/peyrardm/Documents/NLP/KLearn-code/outputs/optimal_K_1_.json', 'r') as f:\n",
    "    optimal_K_1 =  json.loads(f.read())\n",
    "    \n",
    "# with open('../outputs/optimal_K_1_.json', 'r') as f:\n",
    "#     optimal_K_1 =  json.loads(f.read())\n",
    "\n",
    "def prepare_df(optimal):\n",
    "    columns = {}\n",
    "    for c in optimal[0].keys():\n",
    "        columns[c] = [opt_k[c] for opt_k in optimal]\n",
    "        \n",
    "    return pd.DataFrame.from_dict(columns)\n",
    "\n",
    "def build_stats_df(df):\n",
    "    mean_values = df.mean(axis=0)\n",
    "    se_values = df.sem(axis=0)\n",
    "    stats_df = pd.concat([mean_values.to_frame('mean'), se_values.to_frame('se')], axis=1)\n",
    "    stats_df['ci-95'] = 2.5 * stats_df['se']\n",
    "    return stats_df.sort_values(by='mean', ascending=False)  \n",
    "\n",
    "optimal_df = build_stats_df(prepare_df(optimal_K_1))\n",
    "\n",
    "optimal_K = optimal_df['mean'].to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "x,y = [], []\n",
    "nb_samples = 3\n",
    "for nb_avg in range(1, len(annotators_K), 3):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(annotators_K.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(IT.KL(average_K, optimal_K))\n",
    "\n",
    "full_avg = IT.avg([v['K'] for key, v in annotators_K.items()])\n",
    "x.append(len(annotators_K))\n",
    "y.append(np.mean(utils.evaluate_K(dataset, full_avg, 1, 1)))\n",
    "#x.append(len(annotators_K))\n",
    "#y.append(utils.evaluate_K(dataset, full_avg, 1, 1))\n",
    "min_max_scaler = preprocessing.MinMaxScaler()\n",
    "y_scaled = min_max_scaler.fit_transform(np.array(y).reshape(-1, 1))\n",
    "\n",
    "dist_opt_ann = pd.DataFrame.from_dict({'# annotators averaged': x, 'Distance to Optimal': y_scaled[:,0]})\n",
    "\n",
    "x,y = [], []\n",
    "nb_samples = 6\n",
    "for nb_avg in range(1, len(domains_K), 3):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(domains_K.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in domains_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(IT.KL(average_K, optimal_K))\n",
    "\n",
    "#full_avg = IT.avg([K for key, K in Non_News_subsets.items()])\n",
    "#x.append(len(News_subsets))\n",
    "#y.append(pipeline.evaluate_generic_K(dataset, full_avg, 1, 1))\n",
    "\n",
    "min_max_scaler = preprocessing.MinMaxScaler()\n",
    "y_scaled = min_max_scaler.fit_transform(np.array(y).reshape(-1, 1))\n",
    "\n",
    "dist_opt_dom = pd.DataFrame.from_dict({'# domains averaged': x, 'Distance to Optimal': y_scaled[:,0]})\n",
    "\n",
    "x,y = [], []\n",
    "nb_samples = 6\n",
    "for nb_avg in range(1, len(Non_News_subsets), 3):\n",
    "    for _ in range(nb_samples):\n",
    "        selected_keys = random.sample(Non_News_subsets.keys(), nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in domains_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(IT.KL(average_K, optimal_K))\n",
    "\n",
    "#full_avg = IT.avg([K for key, K in Non_News_subsets.items()])\n",
    "#x.append(len(News_subsets))\n",
    "#y.append(pipeline.evaluate_generic_K(dataset, full_avg, 1, 1))\n",
    "\n",
    "min_max_scaler = preprocessing.MinMaxScaler()\n",
    "y_scaled = min_max_scaler.fit_transform(np.array(y).reshape(-1, 1))\n",
    "\n",
    "dist_opt_dom_non_news = pd.DataFrame.from_dict({'# domains averaged': x, 'Distance to Optimal': y_scaled[:,0]})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#sns.set_context('paper')\n",
    "fig, axes = plt.subplots(2, 1, figsize=(8, 10), sharey=False, sharex=False)\n",
    "\n",
    "ax = axes[0]\n",
    "\n",
    "x = averaging_generalizability_ann_test[:-1].groupby('# annotators averaged').mean().index.to_list()\n",
    "y = averaging_generalizability_ann_test[:-1].groupby('# annotators averaged').mean()['Correlation with humans'].to_list()\n",
    "yerr = 1.96*averaging_generalizability_ann_test[:-1].groupby('# annotators averaged').sem()['Correlation with humans'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr, elinewidth=2, linewidth=2, fmt='-x', mew=2, ms=7, color='tab:red')\n",
    "ax.set_ylim(0.0,0.50)\n",
    "\n",
    "#ax.axhline(y=0.2, ls='--', linewidth=2, c='tab:green', label='Baseline')\n",
    "# ax.axhline(y=0.45, ls='--', linewidth=2, c='#34495e', label='hPL')\n",
    "\n",
    "ax.set_xlabel('Number of annotators averaged', fontsize=18)\n",
    "ax.set_ylabel('Correlation with humans', fontsize=18)\n",
    "\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "\n",
    "\n",
    "\n",
    "x = averaging_generalizability_dom.groupby('# domains averaged').mean().index.to_list()\n",
    "y = averaging_generalizability_dom.groupby('# domains averaged').mean()['Correlation with humans'].to_list()\n",
    "yerr = 1.96*averaging_generalizability_dom.groupby('# domains averaged').sem()['Correlation with humans'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr,elinewidth=2, linewidth=2, fmt='-o', ms=7, color='tab:grey')\n",
    "\n",
    "x = averaging_generalizability_non_news.groupby('# domains averaged').mean().index.to_list()\n",
    "y = averaging_generalizability_non_news.groupby('# domains averaged').mean()['Correlation with humans'].to_list()\n",
    "yerr = 1.96*averaging_generalizability_non_news.groupby('# domains averaged').sem()['Correlation with humans'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr,elinewidth=2, linewidth=2, fmt='-s', ms=7, color='tab:blue')\n",
    "\n",
    "ax.set_ylim(0.0,0.50)\n",
    "ax.set_xlim(0,14)\n",
    "\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "\n",
    "ax.axhline(y=0.17, ls='--', linewidth=2, c='tab:green', label='Baseline')\n",
    "ax.axhline(y=0.45, ls='--', linewidth=2, c=\"#34495e\", label='hPL')\n",
    "\n",
    "ax.set_xlabel('Number of annotators/domains averaged', fontsize=18)\n",
    "ax.set_title('a) Correlation improvements', fontsize=23, y=-0.3)\n",
    "#ax.set_ylabel('Correlation with humans', fontsize=13)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "ax = axes[1]\n",
    "\n",
    "\n",
    "\n",
    "x = dist_opt_ann.groupby('# annotators averaged').mean().index.to_list()\n",
    "y = dist_opt_ann.groupby('# annotators averaged').mean()['Distance to Optimal'].to_list()\n",
    "yerr = 2*dist_opt_ann.groupby('# annotators averaged').sem()['Distance to Optimal'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr, elinewidth=2, linewidth=2, fmt='-x', mew=2, ms=7, color='tab:red')\n",
    "\n",
    "\n",
    "ax.set_xlabel('Number of annotators/domains averaged', fontsize=18)\n",
    "ax.set_ylabel('Normalized distance to optimal K', fontsize=18)\n",
    "\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "\n",
    "\n",
    "x = dist_opt_dom.groupby('# domains averaged').mean().index.to_list()\n",
    "y = dist_opt_dom.groupby('# domains averaged').mean()['Distance to Optimal'].to_list()\n",
    "yerr = 2*dist_opt_dom.groupby('# domains averaged').sem()['Distance to Optimal'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr,elinewidth=2, linewidth=2, fmt='-o', ms=7, color='tab:grey')\n",
    "\n",
    "x = dist_opt_dom_non_news.groupby('# domains averaged').mean().index.to_list()\n",
    "y = dist_opt_dom_non_news.groupby('# domains averaged').mean()['Distance to Optimal'].to_list()\n",
    "yerr = 2*dist_opt_dom_non_news.groupby('# domains averaged').sem()['Distance to Optimal'].to_numpy()\n",
    "\n",
    "ax.errorbar(x,y,yerr,elinewidth=2, linewidth=2, fmt='-s', ms=7, color='tab:blue')\n",
    "ax.set_title('b) Reduced distance to optimal', fontsize=23, y=-0.3)\n",
    "\n",
    "\n",
    "#g.legend().set_visible(False)\n",
    "legend_elem = [Line2D([0], [0], linestyle='--', linewidth=2, c='tab:green', label='Baseline (no average)'),\n",
    "               Line2D([0], [0], linestyle='--', linewidth=2, c=\"#34495e\", label='Optimal (HPL)'),\n",
    "               Line2D([0], [0], marker='x', linewidth=2, ms=9, mew=2, c='tab:red', label='Annotators'),\n",
    "               Line2D([0], [0], marker='o', linewidth=2, ms=9, c='tab:grey', label='All domains'),\n",
    "              Line2D([0], [0], marker='s', linewidth=2, ms=9, c='tab:blue', label='Non-News domains')]\n",
    "\n",
    "fig.legend(handles=legend_elem, ncol=2, loc='upper center', frameon=False, fontsize=21,  bbox_to_anchor=(0.55, 1.15))\n",
    "#fig.legend(handles=legend_elem, fontsize=13)\n",
    "fig.tight_layout(pad=1.1)\n",
    "fig.savefig(\"../outputs/averaging_double.pdf\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Annotator Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Averaging improves performances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "TAC08_ann = [ann for ann in annotators_K.keys() if '8' in ann]\n",
    "TAC09_ann = [ann for ann in annotators_K.keys() if '9' in ann]\n",
    "\n",
    "x,y = [], []\n",
    "nb_samples = 3\n",
    "for nb_avg in range(1, int(len(annotators_K) /2.) , 3):\n",
    "    for _ in range(nb_samples):\n",
    "        ## Select from TAC-08 and evaluate on TAC-09\n",
    "        selected_keys = random.sample(TAC08_ann, nb_avg)\n",
    "        average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        x.append(nb_avg)\n",
    "        y.append(np.mean(utils.evaluate_K(dataset_09, average_K, 1, 1)))\n",
    "        \n",
    "        ## Select from TAC-09 and evaluate on TAC-08\n",
    "        #selected_keys = random.sample(TAC09_ann, nb_avg)\n",
    "        #average_K = IT.avg([v['K'] for key, v in annotators_K.items() if key in selected_keys])\n",
    "        \n",
    "        #x.append(nb_avg)\n",
    "        #y.append(np.mean(utils.evaluate_K(dataset_08, average_K, 1, 1)))\n",
    "\n",
    "\n",
    "\n",
    "#full_avg = IT.avg([v['K'] for key, v in annotators_K.items()])\n",
    "#x.append(len(annotators_K))\n",
    "#y.append(np.mean(utils.evaluate_K(dataset, full_avg, 1, 1)))\n",
    "#x.append(len(annotators_K))\n",
    "#y.append(utils.evaluate_K(dataset, full_avg, 1, 1))\n",
    "averaging_generalizability_ann_test = pd.DataFrame.from_dict({'# annotators averaged': x, 'Correlation with humans': y})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plots For Paper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_corr_geom(annotators_K):\n",
    "    keys = annotators_K.keys()\n",
    "    idx = range(len(keys))\n",
    "    keys_idx = dict(zip(keys, idx))\n",
    "    \n",
    "    sizes = []\n",
    "    annots = list(annotators_K.items())\n",
    "    corr = np.ones((len(keys), len(keys)))\n",
    "    for a_i, v_i in annots:\n",
    "        #sizes.append(np.mean(utils.evaluate_K(dataset, v_i['K'], 1., 1.)))\n",
    "        for a_j, v_j in annots:\n",
    "            x_i = keys_idx[a_i]\n",
    "            x_j = keys_idx[a_j]\n",
    "            corr[x_i][x_j] = IT.JS(v_i, v_j)\n",
    "    return corr, keys_idx, np.array(sizes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('../../KLearn-code/outputs/optimal_K_1_.json', 'r') as f:\n",
    "    optimal_K_1 =  json.loads(f.read())\n",
    "    \n",
    "def prepare_df(optimal):\n",
    "    columns = {}\n",
    "    for c in optimal[0].keys():\n",
    "        columns[c] = [opt_k[c] for opt_k in optimal]\n",
    "        \n",
    "    return pd.DataFrame.from_dict(columns)\n",
    "\n",
    "def build_stats_df(df):\n",
    "    mean_values = df.mean(axis=0)\n",
    "    se_values = df.sem(axis=0)\n",
    "    stats_df = pd.concat([mean_values.to_frame('mean'), se_values.to_frame('se')], axis=1)\n",
    "    stats_df['ci-95'] = 2.5 * stats_df['se']\n",
    "    return stats_df.sort_values(by='mean', ascending=False)  \n",
    "\n",
    "optimal_df = build_stats_df(prepare_df(optimal_K_1))\n",
    "optimal_K = optimal_df['mean'].to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.4566147878345657\n",
      "14.493371212121213\n"
     ]
    }
   ],
   "source": [
    "a = np.mean(utils.evaluate_K(dataset_08, optimal_K, 1., 1.))\n",
    "b = np.mean(utils.evaluate_K(dataset_09, optimal_K, 1., 1.))\n",
    "print((a + b)/2)\n",
    "a = np.mean(utils.evaluate_MR(dataset_08, optimal_K, 1., 1.))\n",
    "b = np.mean(utils.evaluate_MR(dataset_09, optimal_K, 1., 1.))\n",
    "print((a + b)/2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('assessor_topics.json', 'r') as f:\n",
    "    assessors = json.loads(f.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "documents_to_keep = [k for k,v in assessors.items() if v == 'E9'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['D0910', 'D0925', 'D0926', 'D0927', 'D0928', 'D0929', 'D0939']\n"
     ]
    }
   ],
   "source": [
    "print(documents_to_keep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "tac2009 = utils.load_data(os.path.join('/Users/peyrardm/Documents/NLP/KLearn-preprocess/data_converted', 'TAC2009_1.pkl'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "text_rep = {}\n",
    "for k,v in tac2009.items():\n",
    "    if k in documents_to_keep:\n",
    "        if k == 'D0925':\n",
    "            continue\n",
    "        text_rep['doc-{}'.format(k)] = v.doc_freq\n",
    "        kept = sorted(v.annotations, key=lambda t: t['pyr_score'], reverse=True)[:4]\n",
    "        for i, summary in enumerate(kept):\n",
    "            text_rep['summary-{}-{}'.format(k, i)] = summary['freq']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "text_rep['K'] = optimal_K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "corr, keys_idx, _ = build_corr_geom(text_rep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "def map_to_type(name):\n",
    "    if 'K' in name:\n",
    "        return 'K'\n",
    "    if 'summary' in name:\n",
    "        return 'summary'\n",
    "    if 'doc' in name:\n",
    "        return 'name'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = corr \n",
    "embedding = MDS(n_components=2, dissimilarity='precomputed')\n",
    "X_transformed = embedding.fit_transform(X)\n",
    "    \n",
    "rep = [f for f, _ in sorted(keys_idx.items(), key=lambda t: t[1])]\n",
    "rep_types = [map_to_type(f) for f in rep]\n",
    "x = X_transformed[:,0]\n",
    "y = X_transformed[:,1]\n",
    "\n",
    "data = pd.DataFrame.from_dict({'X': x, 'Y': y, 'rep':rep, 'rep_types': rep_types})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rand_jitter(arr):\n",
    "    stdev = .01*(max(arr)-min(arr))\n",
    "    return arr + np.random.randn(len(arr)) * stdev\n",
    "\n",
    "\n",
    "data_with_jitter = data.copy()\n",
    "data_with_jitter['X'] = rand_jitter(data_with_jitter['X'])\n",
    "data_with_jitter['Y'] = rand_jitter(data_with_jitter['Y'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Remove jitter from K\n",
    "data_with_jitter.loc[data_with_jitter['rep'] == 'K', 'X'] = 0\n",
    "data_with_jitter.loc[data_with_jitter['rep'] == 'K', 'Y'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 41.0, '')"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.lines as mlines\n",
    "\n",
    "sns.set_context('paper')\n",
    "fig, ax = plt.subplots(1, 1, figsize=(6, 5))\n",
    "\n",
    "\n",
    "#plt.figure(figsize=(6,5))\n",
    "colors = ['tab:blue', 'tab:orange', 'tab:red']\n",
    "sns.scatterplot(x='X', y='Y', ax=ax, hue='rep_types', style=\"rep_types\", \n",
    "                        data=data_with_jitter, s=100, palette=colors, alpha=0.9, legend=False, zorder=120)\n",
    "\n",
    "#for ann, x, y in zip(annotators_K.keys(), x, y):\n",
    "#    ax.text(x-0.018, y-0.01, ann, color = 'white', fontsize=11)\n",
    "    \n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "ax.spines['left'].set_visible(False)\n",
    "ax.spines['bottom'].set_visible(False)\n",
    "\n",
    "ax.grid()\n",
    "#ax.set_xticklabels([])\n",
    "#ax.set_yticklabels([])\n",
    "\n",
    "for tic in ax.xaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "    \n",
    "for tic in ax.yaxis.get_major_ticks():\n",
    "    tic.tick1On = tic.tick2On = False\n",
    "    tic.label1On = tic.label2On = False\n",
    "#ax.get_legend().set_visible(False)\n",
    "\n",
    "# l = mlines.Line2D([0,0.1], [0, 0.5], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "# l = mlines.Line2D([0,0.5], [0, 0.1], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "# l = mlines.Line2D([0,0.4], [0, -0.4], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "# l = mlines.Line2D([0,-0.1], [0, -0.45], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "# l = mlines.Line2D([0,-0.5], [0, -0.1], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "# l = mlines.Line2D([0,-0.4], [0, 0.4], linestyle='--', linewidth=2, color='tab:grey', zorder=10, alpha=0.8)\n",
    "# ax.add_line(l)\n",
    "    \n",
    "legend_elem = [\n",
    "#                 mpatches.Patch(facecolor=colors[0], label='Documents'),\n",
    "#                mpatches.Patch(facecolor=colors[1], label='Reference summaries'),\n",
    "#                mpatches.Patch(facecolor=colors[2], label='K (inferred by hPL)'),\n",
    "               Line2D([0], [0], marker='o', linestyle='', ms=9, c=colors[0], alpha=0.9, label='Documents'),\n",
    "               Line2D([0], [0], marker='x', linestyle='', ms=9, mew=3, c=colors[1], alpha=0.9, label='Reference summaries'),\n",
    "               Line2D([0], [0], marker='s', linestyle='', ms=9, c=colors[2], alpha=0.9, label='K (inferred by HPL)'),\n",
    "               Line2D([0], [0], linestyle='--', linewidth=2, c='tab:grey', alpha=0.9, label='Topic separation')\n",
    "              ]\n",
    "\n",
    "fig.legend(handles=legend_elem, ncol=2, loc='upper center', fontsize=15, bbox_to_anchor=(0.53, 1.1), frameon=False)\n",
    "fig.tight_layout(pad=1)\n",
    "ax.set_ylabel('')\n",
    "ax.set_xlabel('')\n",
    "# fig.savefig(\"../outputs/geometry.pdf\", bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "hide_input": false,
  "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": "3.7.3"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}