Bias_Score_Analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "import matplotlib.pyplot as plt\n",
    "import math\n",
    "stopwords =  set(stopwords.words('english'))\n",
    "import pandas as pd\n",
    "#biased_m_f for m>0 and f>0\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/training_bs/bias_scores/biased_words_m_f') as f :\n",
    "    dm_train = json.load(f)\n",
    "    \n",
    "\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs/bias_scores/biased_words_m_f') as f :\n",
    "    dm_gen = json.load(f)   \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt01/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt01 = json.load(f)    \n",
    "\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt1/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt1 = json.load(f)    \n",
    "    \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt4/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt4 = json.load(f)    \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_1pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_1pt0 = json.load(f)\n",
    "\n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_5pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_5pt0 = json.load(f)\n",
    "\n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_10pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_10pt0 = json.load(f)   \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "U ={}\n",
    "V={}\n",
    "Z ={}\n",
    "for x in dm_train:\n",
    "    if x in dm_gen and x in dm_0pt01 and x in dm_0pt1 and x in dm_0pt4 and x in dm_1pt0 and x in dm_5pt0 and x in dm_10pt0 and x not in stopwords:\n",
    "#         U[x] = { 'training' : abs(0.5 - dm_train[x]['b_score']) , 'generated_text' : abs (0.5 - dm_gen[x]['b_score']), 'l_0pt4' : abs (0.5 - dm_0pt4[x]['b_score']) ,'l_1pt0' : abs (0.5 - dm_1pt0[x]['b_score']), 'l_5pt0' : abs (0.5 - dm_5pt0[x]['b_score']), 'l_10pt0' : abs (0.5 - dm_10pt0[x]['b_score'])}\n",
    "        #W[x] = { 'training' : dm_train[x]['b_score']) , 'generated_text' : abs (0.5 - dm_gen[x]['b_score']), 'l_0pt4' : abs (0.5 - dm_0pt4[x]['b_score']) ,'l_1pt0' : abs (0.5 - dm_1pt0[x]['b_score']), 'l_5pt0' : abs (0.5 - dm_5pt0[x]['b_score']), 'l_10pt0' : abs (0.5 - dm_10pt0[x]['b_score'])}\n",
    "        \n",
    "        V[x] = { 'training' :2*(0.5 - dm_train[x]['b_score']) , 'generated_text' : 2* (0.5 - dm_gen[x]['b_score']) ,'l_0pt01' : 2* (0.5 - dm_0pt01[x]['b_score']), 'l_0pt1' : 2* (0.5 - dm_0pt1[x]['b_score']),'l_0pt4' : 2* (0.5 - dm_0pt4[x]['b_score']) ,'l_1pt0' : 2* (0.5 - dm_1pt0[x]['b_score']), 'l_5pt0' : 2* (0.5 - dm_5pt0[x]['b_score']), 'l_10pt0' : 2* (0.5 - dm_10pt0[x]['b_score'])}\n",
    "        Z[x] = { 'training' :(dm_train[x]['b_score']) , 'generated_text' :  ( dm_gen[x]['b_score']) ,'l_0pt01' :  (dm_0pt01[x]['b_score']), 'l_0pt1' :  (dm_0pt1[x]['b_score']),'l_0pt4' :  (dm_0pt4[x]['b_score']) ,'l_1pt0' :dm_1pt0[x]['b_score'], 'l_5pt0' :  dm_5pt0[x]['b_score'], 'l_10pt0' : ( dm_10pt0[x]['b_score'])}\n",
    "        #print(x)\n",
    "        \n",
    "# M = pd.DataFrame.from_dict(U, orient='index')   \n",
    "N = pd.DataFrame.from_dict(V, orient='index')   \n",
    "L = pd.DataFrame.from_dict(Z, orient='index')  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2ab47b029828>"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2ab482dd6c18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "L.columns.name = ['training', 'generated text', 'l_0pt01', 'l_0pt1','l_0pt4', 'l_1pt0', 'l_10pt0', 'l_10pt0'] \n",
    "#plt = L.plot(columns=['Type A', 'Type B'], index=['Index 1','Index 2']))\n",
    "L.training.plot.kde(legend = True, y = 'frequency')\n",
    "L.generated_text.plot.kde(legend = True)\n",
    "L.l_0pt01.plot.kde(legend = True)\n",
    "L.l_0pt1.plot.kde(legend = True)\n",
    "L.l_0pt4.plot.kde(legend = True)\n",
    "L.l_1pt0.plot.kde(legend = True)\n",
    "L.l_5pt0.plot.kde(legend = True)\n",
    "L.l_10pt0.plot.kde(legend = True)\n",
    "\n",
    "# print (L.training.mean())\n",
    "# print (L.generated_text.mean())\n",
    "# print (L.l_0pt01.mean())\n",
    "# print (L.l_0pt1.mean())\n",
    "# print (L.l_0pt4.mean())\n",
    "# print (L.l_1pt0.mean())\n",
    "# print(L.l_5pt0.mean())\n",
    "# print (L.l_10pt0.mean())\n",
    "# print (L.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# M.columns.name = ['training', 'generated text',  'l_0pt01',  'l_0pt1', 'l_0pt4', 'l_1pt0', 'l_10pt0', 'l_10pt0'] \n",
    "# #plt = L.plot(columns=['Type A', 'Type B'], index=['Index 1','Index 2']))\n",
    "# M.training.plot.kde(legend = True, y = 'frequency')\n",
    "# M.generated_text.plot.kde(legend = True)\n",
    "# M.l_0pt01.plot.kde(legend = True)\n",
    "# M.l_0pt1.plot.kde(legend = True)\n",
    "# M.l_0pt4.plot.kde(legend = True)\n",
    "# M.l_1pt0.plot.kde(legend = True)\n",
    "# M.l_5pt0.plot.kde(legend = True)\n",
    "# M.l_10pt0.plot.kde(legend = True)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "Mean=[]\n",
    "Median = []\n",
    "Std_dev = []\n",
    "Skew =[]\n",
    "Kurtosis =[]\n",
    "Lambda =np.array([0, 0, 0.01,0.1,0.4,1.0, 5.0, 10.0 ], dtype =float)\n",
    "# 257,255.33, 258.46, 244.28,238.91, 238.91, 238.96\n",
    "Perplexity = np.array([0, 257,255.33, 258.46, 244.28,238.91, 238.91, 238.96])\n",
    "Row_names=['training', 'generated text', 'lambda_0pt01','lambda_0pt1','lambda_0pt4', 'lambda_1pt0', 'lambda_5pt0','lambda_10.0']\n",
    "for column in N.columns:\n",
    "    Mean.append(round(N[column].mean(),2))\n",
    "    Median.append(round(N[column].median(),2))\n",
    "                \n",
    "    Std_dev.append(round(N[column].std(),2))\n",
    "    Skew.append(round(N[column].skew(),2))\n",
    "    Kurtosis.append(round(N[column].kurtosis(),2))\n",
    "\n",
    "\n",
    "stats = pd.DataFrame(np.column_stack([Row_names,Lambda,Perplexity, Mean, Median, Std_dev, Skew, Kurtosis]), \n",
    "                               columns=['label', 'lambda','Perplexity','Mean', 'Median', 'Standard Deviation','Skew','Kurtosis'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>lambda</th>\n",
       "      <th>Perplexity</th>\n",
       "      <th>Mean</th>\n",
       "      <th>Median</th>\n",
       "      <th>Standard Deviation</th>\n",
       "      <th>Skew</th>\n",
       "      <th>Kurtosis</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>training</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.22</td>\n",
       "      <td>-0.24</td>\n",
       "      <td>0.24</td>\n",
       "      <td>0.94</td>\n",
       "      <td>2.44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>generated text</td>\n",
       "      <td>0.0</td>\n",
       "      <td>257.0</td>\n",
       "      <td>-0.28</td>\n",
       "      <td>-0.31</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.68</td>\n",
       "      <td>0.69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>lambda_0pt01</td>\n",
       "      <td>0.01</td>\n",
       "      <td>255.33</td>\n",
       "      <td>-0.32</td>\n",
       "      <td>-0.33</td>\n",
       "      <td>0.28</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>lambda_0pt1</td>\n",
       "      <td>0.1</td>\n",
       "      <td>258.46</td>\n",
       "      <td>-0.3</td>\n",
       "      <td>-0.33</td>\n",
       "      <td>0.3</td>\n",
       "      <td>0.77</td>\n",
       "      <td>0.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>lambda_0pt4</td>\n",
       "      <td>0.4</td>\n",
       "      <td>244.28</td>\n",
       "      <td>-0.35</td>\n",
       "      <td>-0.39</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.89</td>\n",
       "      <td>1.02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>lambda_1pt0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>238.91</td>\n",
       "      <td>-0.3</td>\n",
       "      <td>-0.33</td>\n",
       "      <td>0.32</td>\n",
       "      <td>0.75</td>\n",
       "      <td>0.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>lambda_5pt0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>238.91</td>\n",
       "      <td>-0.35</td>\n",
       "      <td>-0.4</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.92</td>\n",
       "      <td>0.97</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>lambda_10.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>238.96</td>\n",
       "      <td>-0.35</td>\n",
       "      <td>-0.4</td>\n",
       "      <td>0.31</td>\n",
       "      <td>0.92</td>\n",
       "      <td>0.97</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            label lambda Perplexity   Mean Median Standard Deviation  Skew  \\\n",
       "0        training    0.0        0.0  -0.22  -0.24               0.24  0.94   \n",
       "1  generated text    0.0      257.0  -0.28  -0.31                0.3  0.68   \n",
       "2    lambda_0pt01   0.01     255.33  -0.32  -0.33               0.28   0.8   \n",
       "3     lambda_0pt1    0.1     258.46   -0.3  -0.33                0.3  0.77   \n",
       "4     lambda_0pt4    0.4     244.28  -0.35  -0.39               0.31  0.89   \n",
       "5     lambda_1pt0    1.0     238.91   -0.3  -0.33               0.32  0.75   \n",
       "6     lambda_5pt0    5.0     238.91  -0.35   -0.4               0.31  0.92   \n",
       "7     lambda_10.0   10.0     238.96  -0.35   -0.4               0.31  0.92   \n",
       "\n",
       "  Kurtosis  \n",
       "0     2.44  \n",
       "1     0.69  \n",
       "2     1.08  \n",
       "3      0.9  \n",
       "4     1.02  \n",
       "5      0.6  \n",
       "6     0.97  \n",
       "7     0.97  "
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "import matplotlib.pyplot as plt\n",
    "stopwords =  set(stopwords.words('english'))\n",
    "import pandas as pd\n",
    "#biased_m_f for m>0 and f>0\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/training_bs/bias_scores/biased_words_m_f') as f :\n",
    "    dm_train = json.load(f)\n",
    "    \n",
    "\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs/bias_scores/biased_words_m_f') as f :\n",
    "    dm_gen = json.load(f)   \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt01/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt01 = json.load(f)    \n",
    "\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt1/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt1 = json.load(f)    \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_0pt4/bias_scores/biased_words_m_f') as f :\n",
    "    dm_0pt4 = json.load(f)    \n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_1pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_1pt0 = json.load(f)\n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_5pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_5pt0 = json.load(f)\n",
    "\n",
    "    \n",
    "with open ('/beegfs/sb6416/ds1012/scores/dm/finale/bs_10pt0/bias_scores/biased_words_m_f') as f :\n",
    "    dm_10pt0 = json.load(f)   \n",
    "    \n",
    "U ={}\n",
    "V={}\n",
    "Z ={}\n",
    "D={}\n",
    "C={}\n",
    "for x in dm_train:\n",
    "    if x in dm_gen and x in dm_0pt01 and x in dm_0pt1 and x in dm_0pt4 and x in dm_1pt0 and x in dm_5pt0 and x in dm_10pt0 and x not in stopwords:\n",
    "        #print(x )\n",
    "#         U[x] = { 'training' : abs(0.5 - dm_train[x]['b_score']) , 'generated_text' : abs (0.5 - dm_gen[x]['b_score']), 'l_0pt4' : abs (0.5 - dm_0pt4[x]['b_score']) ,'l_1pt0' : abs (0.5 - dm_1pt0[x]['b_score']), 'l_5pt0' : abs (0.5 - dm_5pt0[x]['b_score']), 'l_10pt0' : abs (0.5 - dm_10pt0[x]['b_score'])}\n",
    "        \n",
    "         V[x] = {'training' :2*(0.5 - dm_train[x]['b_score']) , \n",
    "                 'generated_text' : 2* (0.5 - dm_gen[x]['b_score']) ,\n",
    "                 'l_0pt01' : 2* (0.5 - dm_0pt01[x]['b_score']),\n",
    "                'l_0pt1' : 2* (0.5 - dm_0pt1[x]['b_score']),\n",
    "                'l_0pt4' : 2* (0.5 - dm_0pt4[x]['b_score']) ,\n",
    "                'l_1pt0' : 2* (0.5 - dm_1pt0[x]['b_score']), \n",
    "                'l_5pt0' : 2* (0.5 - dm_5pt0[x]['b_score']),\n",
    "                'l_10pt0' : 2* (0.5 - dm_10pt0[x]['b_score'])}\n",
    "        \n",
    "#         Z[x] = { 'training' :(dm_train[x]['b_score']) , 'generated_text' :  ( dm_gen[x]['b_score']) ,'l_0pt4' :  (dm_0pt4[x]['b_score']) ,'l_1pt0' :dm_1pt0[x]['b_score'], 'l_5pt0' :  dm_5pt0[x]['b_score'], 'l_10pt0' : ( dm_10pt0[x]['b_score'])}\n",
    "         #D[x]=  { 'gen' :2*((0.5 - dm_train[x]['b_score']) - (0.5- dm_gen[x]['b_score'])), 'l_0pt01' :2*((0.5 - dm_train[x]['b_score']) - (0.5 - dm_0pt01[x]['b_score'])),'l_0pt1' :2*((0.5 - dm_train[x]['b_score']) - ( dm_0pt1[x]['b_score']),'l_0pt4' :(dm_train[x]['b_score']) - ( dm_0pt4[x]['b_score']),   'l_1pt0' :(dm_train[x]['b_score']) - ( dm_1pt0[x]['b_score']),  'l_5pt0' :(dm_train[x]['b_score']) - ( dm_5pt0[x]['b_score']), 'l_10pt0' :(dm_train[x]['b_score']) - (0.5 -  dm_10pt0[x]['b_score']), }\n",
    "         \n",
    "         C[x]=  { 'gen' :abs(dm_train[x]['b_score'] -  dm_gen[x]['b_score']),'l_0pt01' :abs(dm_train[x]['b_score'] -  dm_0pt01[x]['b_score']),'l_0pt1' :abs(dm_train[x]['b_score'] -  dm_0pt1[x]['b_score']),'l_0pt4' :abs(dm_train[x]['b_score'] -  dm_0pt4[x]['b_score']),   'l_1pt0' :abs(dm_train[x]['b_score'] -  dm_1pt0[x]['b_score']),  'l_5pt0' :abs(dm_train[x]['b_score'] -  dm_5pt0[x]['b_score']), 'l_10pt0' :abs(dm_train[x]['b_score'] - dm_10pt0[x]['b_score']), }\n",
    "        #print(x)\n",
    "        \n",
    "# M = pd.DataFrame.from_dict(U, orient='index')   \n",
    "# N = pd.DataFrame.from_dict(V, orient='index')   \n",
    "# L = pd.DataFrame.from_dict(Z, orient='index')  \n",
    "\n",
    "for x in V:\n",
    "    D[x]={'delta_generated_text' : abs(V[x]['training'] - V[x]['generated_text']), \n",
    "          'delta_lambda_0pt01' : abs(V[x]['training'] - V[x]['l_0pt01']), \n",
    "          'delta_lambda_0pt1' : abs(V[x]['training'] - V[x]['l_0pt1']), \n",
    "          'delta_lambda_0pt4' : abs(V[x]['training'] - V[x]['l_0pt4']), \n",
    "          'delta_lambda_1pt0' : abs(V[x]['training'] - V[x]['l_1pt0']), \n",
    "          'delta_lambda_5pt0' : abs(V[x]['training'] - V[x]['l_5pt0']),\n",
    "          'delta_lambda_10pt0' : abs(V[x]['training'] - V[x]['l_10pt0'])}\n",
    "\n",
    "X = pd.DataFrame.from_dict(D, orient='index')\n",
    "Y = pd.DataFrame.from_dict(C, orient='index')\n",
    "\n",
    "\n",
    "# L.columns.name = ['training', 'generated text', 'l_0pt01', 'l_0pt1','l_0pt4', 'l_1pt0', 'l_10pt0', 'l_10pt0'] \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2b85ef3f4d30>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2b85f0e1c898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X.delta_generated_text.plot.kde(legend = True)\n",
    "X.delta_lambda_0pt01.plot.kde(legend = True)\n",
    "X.delta_lambda_0pt1.plot.kde(legend = True)\n",
    "X.delta_lambda_0pt4.plot.kde(legend = True)\n",
    "X.delta_lambda_1pt0.plot.kde(legend = True)\n",
    "X.delta_lambda_5pt0.plot.kde(legend = True)\n",
    "X.delta_lambda_10pt0.plot.kde(legend = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "Mean=[]\n",
    "Median = []\n",
    "Std_dev = []\n",
    "Skew =[]\n",
    "Kurtosis =[]\n",
    "Lambda =np.array([ 0,0.01, 0.1,  0.4,1.0, 5.0, 10.0 ], dtype =float)\n",
    "# 257,255.33, 258.46, 244.28,238.91, 238.91, 238.96\n",
    "Perplexity = np.array([257,255.33, 258.46, 244.28,238.91, 238.91, 238.96])\n",
    "Row_names=[ 'generated text', 'lambda_0pt01', 'lambda_0pt1','lambda_0pt4', 'lambda_1pt0', 'lambda_5pt0','lambda_10.0']\n",
    "for column in X.columns:\n",
    "    Mean.append(round(X[column].mean(),4))\n",
    "#Median.append(round(Y[column].median(),4))\n",
    "                  \n",
    "   # Std_dev.append(round(Y[column].std(),4))\n",
    "    Skew.append(round(X[column].skew(),4))\n",
    "   # Kurtosis.append(round(Y[column].kurtosis(),4))\n",
    "\n",
    "\n",
    "stats = pd.DataFrame(np.column_stack([Lambda,Perplexity, Mean,Skew]), \n",
    "                               columns=[ 'lambda','Perplexity','Mean', 'Skew'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lambda</th>\n",
       "      <th>Perplexity</th>\n",
       "      <th>Mean</th>\n",
       "      <th>Skew</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.00</td>\n",
       "      <td>257.00</td>\n",
       "      <td>0.2275</td>\n",
       "      <td>1.4665</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.01</td>\n",
       "      <td>255.33</td>\n",
       "      <td>0.2366</td>\n",
       "      <td>1.3873</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.10</td>\n",
       "      <td>258.46</td>\n",
       "      <td>0.2328</td>\n",
       "      <td>1.3903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.40</td>\n",
       "      <td>244.28</td>\n",
       "      <td>0.2550</td>\n",
       "      <td>1.2193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1.00</td>\n",
       "      <td>238.91</td>\n",
       "      <td>0.2408</td>\n",
       "      <td>1.2646</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5.00</td>\n",
       "      <td>238.91</td>\n",
       "      <td>0.2615</td>\n",
       "      <td>1.1762</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>10.00</td>\n",
       "      <td>238.96</td>\n",
       "      <td>0.2615</td>\n",
       "      <td>1.1762</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   lambda  Perplexity    Mean    Skew\n",
       "0    0.00      257.00  0.2275  1.4665\n",
       "1    0.01      255.33  0.2366  1.3873\n",
       "2    0.10      258.46  0.2328  1.3903\n",
       "3    0.40      244.28  0.2550  1.2193\n",
       "4    1.00      238.91  0.2408  1.2646\n",
       "5    5.00      238.91  0.2615  1.1762\n",
       "6   10.00      238.96  0.2615  1.1762"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stats"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}