create_post_engage.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from parlai.core.agents import create_agent\n",
    "from parlai.core.worlds import create_task\n",
    "from parlai.core.image_featurizers import ImageLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cgi\n",
    "import PIL.Image as IM\n",
    "from base64 import b64decode\n",
    "import io\n",
    "import os\n",
    "import base64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from os import walk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "import spacy\n",
    "import pandas as pd\n",
    "from nearpy.distances import CosineDistance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Paths dev"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# captioning\n",
    "opt_download_path =''\n",
    "opt_datapath=''\n",
    "opt_parlai_home=''\n",
    "opt_model_file = './transresnet/model'\n",
    "opt_fixed_cands_path='./transresnet/candidates.txt'\n",
    "personality_captions_path='./personality_captions/train.json'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Captioning utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt={'init_opt': None, \n",
    "     'show_advanced_args': False, \n",
    "     'task': 'interactive',\n",
    "     'download_path': opt_download_path, \n",
    "     'datatype': 'train',\n",
    "     'image_mode': 'raw', \n",
    "     'numthreads': 1,\n",
    "     'hide_labels': False,\n",
    "     'multitask_weights': [1],\n",
    "     'batchsize': 1,\n",
    "     'datapath': opt_datapath, \n",
    "     'model': None, \n",
    "     'model_file': opt_model_file, \n",
    "     'init_model': None, \n",
    "     'dict_class': None, \n",
    "     'display_examples': False, \n",
    "     'display_prettify': False, \n",
    "     'display_ignore_fields': \n",
    "     'label_candidates,text_candidates', \n",
    "     'interactive_task': True, \n",
    "     'local_human_candidates_file': None,\n",
    "     'single_turn': False,\n",
    "     'image_size': 256,\n",
    "     'image_cropsize': 224,\n",
    "     'embedding_size': 300, \n",
    "     'n_layers': 2, \n",
    "     'ffn_size': 1200, \n",
    "     'dropout': 0.0, \n",
    "     'attention_dropout': 0.2,\n",
    "     'relu_dropout': 0.2, \n",
    "     'n_heads': 2,\n",
    "     'learn_positional_embeddings': False,\n",
    "     'embeddings_scale': True,\n",
    "     'n_positions': 1000, \n",
    "     'n_segments': 0,\n",
    "     'variant': 'aiayn',\n",
    "     'activation': 'relu',\n",
    "     'output_scaling': 1.0, \n",
    "     'share_word_embeddings': True,\n",
    "     'truncate': 32, \n",
    "     'image_features_dim': 2048,\n",
    "     'embedding_type': None,\n",
    "     'load_encoder_from': None, \n",
    "     'hidden_dim': 300, \n",
    "     'num_layers_all': -1,\n",
    "     'num_layers_text_encoder': 1,\n",
    "     'num_layers_image_encoder': 1,\n",
    "     'no_cuda': False, \n",
    "     'learningrate': 0.0005,\n",
    "     'additional_layer_dropout': 0.2, \n",
    "     'freeze_patience': -1, \n",
    "     'one_cand_set': False, \n",
    "     'fixed_cands_path': None,\n",
    "     'pretrained': False, \n",
    "     'dict_file': None,\n",
    "     'dict_initpath': None, \n",
    "     'dict_language': 'english',\n",
    "     'dict_max_ngram_size': -1,\n",
    "     'dict_minfreq': 0, \n",
    "     'dict_maxtokens': -1, \n",
    "     'dict_nulltoken': '__null__',\n",
    "     'dict_starttoken': '__start__', \n",
    "     'dict_endtoken': '__end__',\n",
    "     'dict_unktoken': '__unk__',\n",
    "     'dict_tokenizer': 're',\n",
    "     'dict_lower': False, \n",
    "     'bpe_debug': False, \n",
    "     'dict_textfields': 'text,labels',\n",
    "     'interactive_mode': True, \n",
    "     'parlai_home': opt_parlai_home, \n",
    "     'override': {'ffn_size': 1200,\n",
    "                  'attention_dropout': 0.2,\n",
    "                  'relu_dropout': 0.2, \n",
    "                  'n_positions': 1000,\n",
    "                  'model_file': opt_model_file},\n",
    "                 'starttime': 'Aug20_13-01'}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt['fixed_cands_path'] = opt_fixed_cands_path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt['task'] = 'parlai.agents.local_human.local_human:LocalHumanAgent'\n",
    "opt['image_mode'] = 'resnet152'\n",
    "opt['no_cuda'] = True\n",
    "opt['override']['no_cuda'] = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/georgy/anaconda3/lib/python3.6/site-packages/torchvision/transforms/transforms.py:188: UserWarning: The use of the transforms.Scale transform is deprecated, please use transforms.Resize instead.\n",
      "  \"please use transforms.Resize instead.\")\n",
      "Downloading: \"https://download.pytorch.org/models/resnet152-b121ed2d.pth\" to /home/georgy/.torch/models/resnet152-b121ed2d.pth\n",
      "241530880it [00:22, 10962379.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ warning: overriding opt['attention_dropout'] to 0.2 (previously: None )]\n",
      "[ warning: overriding opt['relu_dropout'] to 0.2 (previously: None )]\n",
      "[ warning: overriding opt['n_positions'] to 1000 (previously: None )]\n",
      "[ warning: overriding opt['model_file'] to /home/georgy/Desktop/Movista/MP/transfer/engaging_social_media_posts/transresnet/model (previously: pers_cap_model_best_resnet )]\n",
      "[ warning: overriding opt['no_cuda'] to True (previously: False )]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/georgy/ParlAI/parlai/agents/transformer/modules.py:32: UserWarning: Installing APEX can give a significant speed boost.\n",
      "  warn_once(\"Installing APEX can give a significant speed boost.\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dictionary: loading dictionary from /home/georgy/Desktop/Movista/MP/transfer/engaging_social_media_posts/transresnet/model.dict\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "Downloading personality_captions.tgz: 0.00B [00:00, ?B/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ num words =  250006 ]\n",
      "[building data: personality_captions]\n",
      "[ downloading: http://parl.ai/downloads/personality_captions/personality_captions.tgz to personality_captions/personality_captions.tgz ]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Downloading personality_captions.tgz: 100%|██████████| 138M/138M [00:15<00:00, 8.79MB/s]   \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unpacking personality_captions.tgz\n",
      "Creating or loading model\n",
      "loading saved cand encodings\n",
      "[creating task(s): parlai.agents.local_human.local_human:LocalHumanAgent]\n",
      "Enter [DONE] if you want to end the episode.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "SHARED = {}\n",
    "SHARED['opt'] = opt\n",
    "SHARED['image_loader'] = ImageLoader(opt)\n",
    "# Create model and assign it to the specified task\n",
    "SHARED['agent'] = create_agent(opt, requireModelExists=True)\n",
    "SHARED['world'] = create_task(opt, SHARED['agent'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Personalities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "persona_dic = {0: 'Abrasive (Annoying, Irritating)',\n",
    " 1: 'Absentminded',\n",
    " 2: 'Adventurous',\n",
    " 3: 'Aggressive',\n",
    " 4: 'Airy (Casual, Not Serious)',\n",
    " 5: 'Aloof (Detached, Distant)',\n",
    " 6: 'Amusing',\n",
    " 7: 'Angry',\n",
    " 8: 'Anxious',\n",
    " 9: 'Apathetic (Uncaring, Disinterested)',\n",
    " 10: 'Appreciative (Grateful)',\n",
    " 11: 'Argumentative',\n",
    " 12: 'Arrogant',\n",
    " 13: 'Artful',\n",
    " 14: 'Articulate (Well-spoken, Expressive)',\n",
    " 15: 'Artificial',\n",
    " 16: 'Assertive',\n",
    " 17: 'Attractive',\n",
    " 18: 'Barbaric',\n",
    " 19: 'Bewildered (Astonished, Confused)',\n",
    " 20: 'Bizarre',\n",
    " 21: 'Bland',\n",
    " 22: 'Blunt',\n",
    " 23: 'Boisterous (Rowdy, Loud)',\n",
    " 24: 'Boyish',\n",
    " 25: 'Breezy (Relaxed, Informal)',\n",
    " 26: 'Brilliant',\n",
    " 27: 'Businesslike',\n",
    " 28: 'Calm',\n",
    " 29: 'Captivating',\n",
    " 30: 'Caring',\n",
    " 31: 'Casual',\n",
    " 32: 'Cerebral (Intellectual, Logical)',\n",
    " 33: 'Charming',\n",
    " 34: 'Cheerful',\n",
    " 35: 'Childish',\n",
    " 36: 'Clever',\n",
    " 37: 'Coarse (Not Fine, Crass)',\n",
    " 38: 'Cold',\n",
    " 39: 'Colorful (Full of Life, Interesting)',\n",
    " 40: 'Compassionate (Sympathetic, Warm)',\n",
    " 41: 'Complex',\n",
    " 42: 'Conceited (Arrogant, Egotistical)',\n",
    " 43: 'Confident',\n",
    " 44: 'Confused',\n",
    " 45: 'Conservative (Traditional, Conventional)',\n",
    " 46: 'Considerate',\n",
    " 47: 'Contemplative (Reflective, Thoughtful)',\n",
    " 48: 'Contemptible (Despicable, Vile)',\n",
    " 49: 'Contradictory',\n",
    " 50: 'Courageous',\n",
    " 51: 'Cowardly',\n",
    " 52: 'Crazy',\n",
    " 53: 'Creative',\n",
    " 54: 'Critical',\n",
    " 55: 'Crude',\n",
    " 56: 'Cruel',\n",
    " 57: 'Cultured (Refined, Educated)',\n",
    " 58: 'Curious',\n",
    " 59: 'Cute',\n",
    " 60: 'Cynical (Doubtful, Skeptical)',\n",
    " 61: 'Daring',\n",
    " 62: 'Deep',\n",
    " 63: 'Destructive',\n",
    " 64: 'Devious',\n",
    " 65: 'Discouraging',\n",
    " 66: 'Disturbing',\n",
    " 67: 'Dramatic',\n",
    " 68: 'Dreamy',\n",
    " 69: 'Dry',\n",
    " 70: 'Dull',\n",
    " 71: 'Earnest',\n",
    " 72: 'Earnest (Enthusiastic)',\n",
    " 73: 'Egocentric (Self-centered)',\n",
    " 74: 'Elegant',\n",
    " 75: 'Eloquent (Well-spoken, Expressive)',\n",
    " 76: 'Emotional',\n",
    " 77: 'Empathetic',\n",
    " 78: 'Energetic',\n",
    " 79: 'Enigmatic (Cryptic, Obscure)',\n",
    " 80: 'Enthusiastic',\n",
    " 81: 'Envious',\n",
    " 82: 'Erratic',\n",
    " 83: 'Escapist (Dreamer, Seeks Distraction)',\n",
    " 84: 'Excitable',\n",
    " 85: 'Exciting',\n",
    " 86: 'Extraordinary',\n",
    " 87: 'Extravagant',\n",
    " 88: 'Extreme',\n",
    " 89: 'Fanatical',\n",
    " 90: 'Fanciful',\n",
    " 91: 'Fatalistic (Bleak, Gloomy)',\n",
    " 92: 'Fawning (Flattering, Deferential)',\n",
    " 93: 'Fearful',\n",
    " 94: 'Fickle (Changeable, Temperamental)',\n",
    " 95: 'Fiery',\n",
    " 96: 'Foolish',\n",
    " 97: 'Formal',\n",
    " 98: 'Freethinking',\n",
    " 99: 'Frightening',\n",
    " 100: 'Frivolous (Trivial, Silly)',\n",
    " 101: 'Fun-loving',\n",
    " 102: 'Gentle',\n",
    " 103: 'Glamorous',\n",
    " 104: 'Gloomy',\n",
    " 105: 'Grand',\n",
    " 106: 'Grim',\n",
    " 107: 'Happy',\n",
    " 108: 'Hateful',\n",
    " 109: 'Haughty (Arrogant, Snobbish)',\n",
    " 110: 'High-spirited',\n",
    " 111: 'Honest',\n",
    " 112: 'Hostile',\n",
    " 113: 'Humble',\n",
    " 114: 'Humorous',\n",
    " 115: 'Idealistic',\n",
    " 116: 'Imaginative',\n",
    " 117: 'Impersonal',\n",
    " 118: 'Insightful',\n",
    " 119: 'Intelligent',\n",
    " 120: 'Intense',\n",
    " 121: 'Irrational',\n",
    " 122: 'Irritable',\n",
    " 123: 'Kind',\n",
    " 124: 'Knowledgeable',\n",
    " 125: 'Lazy',\n",
    " 126: 'Logical',\n",
    " 127: 'Malicious',\n",
    " 128: 'Maternal (Mother-like)',\n",
    " 129: 'Melancholic',\n",
    " 130: 'Mellow (Soothing, Sweet)',\n",
    " 131: 'Meticulous (Precise, Thorough)',\n",
    " 132: 'Miserable',\n",
    " 133: 'Money-minded',\n",
    " 134: 'Monstrous',\n",
    " 135: 'Moody',\n",
    " 136: 'Morbid',\n",
    " 137: 'Mystical',\n",
    " 138: 'Narcissistic (Self-centered, Egotistical)',\n",
    " 139: 'Neurotic (Manic, Obsessive)',\n",
    " 140: 'Neutral',\n",
    " 141: 'Nihilistic',\n",
    " 142: 'Objective (Detached, Impartial)',\n",
    " 143: 'Obnoxious',\n",
    " 144: 'Observant',\n",
    " 145: 'Obsessive',\n",
    " 146: 'Odd',\n",
    " 147: 'Offhand',\n",
    " 148: 'Old-fashioned',\n",
    " 149: 'Open',\n",
    " 150: 'Opinionated',\n",
    " 151: 'Optimistic',\n",
    " 152: 'Ordinary',\n",
    " 153: 'Outrageous',\n",
    " 154: 'Overimaginative',\n",
    " 155: 'Paranoid',\n",
    " 156: 'Passionate',\n",
    " 157: 'Passive',\n",
    " 158: 'Patriotic',\n",
    " 159: 'Peaceful',\n",
    " 160: 'Perceptive',\n",
    " 161: 'Playful',\n",
    " 162: 'Pompous (Self-important, Arrogant)',\n",
    " 163: 'Practical',\n",
    " 164: 'Pretentious (Snobbish, Showy)',\n",
    " 165: 'Profound',\n",
    " 166: 'Provocative',\n",
    " 167: 'Questioning',\n",
    " 168: 'Quirky',\n",
    " 169: 'Rational',\n",
    " 170: 'Realistic',\n",
    " 171: 'Reflective',\n",
    " 172: 'Relaxed',\n",
    " 173: 'Resentful',\n",
    " 174: 'Respectful',\n",
    " 175: 'Ridiculous',\n",
    " 176: 'Rigid',\n",
    " 177: 'Romantic',\n",
    " 178: 'Rowdy',\n",
    " 179: 'Rustic (Rural)',\n",
    " 180: 'Sarcastic',\n",
    " 181: 'Scholarly',\n",
    " 182: 'Scornful',\n",
    " 183: 'Sensitive',\n",
    " 184: 'Sensual',\n",
    " 185: 'Sentimental',\n",
    " 186: 'Serious',\n",
    " 187: 'Shy',\n",
    " 188: 'Silly',\n",
    " 189: 'Simple',\n",
    " 190: 'Skeptical',\n",
    " 191: 'Solemn',\n",
    " 192: 'Sophisticated',\n",
    " 193: 'Spirited',\n",
    " 194: 'Spontaneous',\n",
    " 195: 'Stiff',\n",
    " 196: 'Stoic (Unemotional, Matter-of-fact)',\n",
    " 197: 'Stupid',\n",
    " 198: 'Stylish',\n",
    " 199: 'Suave (Charming, Smooth)',\n",
    " 200: 'Sweet',\n",
    " 201: 'Sympathetic',\n",
    " 202: 'Tense',\n",
    " 203: 'Tough',\n",
    " 204: 'Uncreative',\n",
    " 205: 'Unimaginative',\n",
    " 206: 'Unrealistic',\n",
    " 207: 'Vacuous (Empty, Unintelligent)',\n",
    " 208: 'Vague',\n",
    " 209: 'Vivacious (Lively, Animated)',\n",
    " 210: 'Warm',\n",
    " 211: 'Whimsical (Playful, Fanciful)',\n",
    " 212: 'Wise',\n",
    " 213: 'Wishful',\n",
    " 214: 'Witty',\n",
    " 215: 'Youthful',\n",
    " 216: 'Zany'}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_json(personality_captions_path)"
   ]
  },
  {
   "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>personality</th>\n",
       "      <th>comment</th>\n",
       "      <th>image_hash</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>Intense</td>\n",
       "      <td>The snow will last as long as my sadness</td>\n",
       "      <td>1e22a9cf867d718551386b427c3b6d18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>Adventurous</td>\n",
       "      <td>I love experiencing new cultures</td>\n",
       "      <td>96472caea58db27769f1c282e2ac0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>Mellow (Soothing, Sweet)</td>\n",
       "      <td>Look at that smooth easy catch of the ball. li...</td>\n",
       "      <td>f09d8fb76822158de129acb0fef463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>Zany</td>\n",
       "      <td>I wish I could just run down this shore!</td>\n",
       "      <td>6e4ccc739ff44ed11da20ad9892317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>Narcissistic (Self-centered, Egotistical)</td>\n",
       "      <td>Psh, I've seen way better floral arrangements....</td>\n",
       "      <td>e7e1844aa9e67cddc6ffe8804d76e45b</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 personality  \\\n",
       "0                                    Intense   \n",
       "1                                Adventurous   \n",
       "2                   Mellow (Soothing, Sweet)   \n",
       "3                                       Zany   \n",
       "4  Narcissistic (Self-centered, Egotistical)   \n",
       "\n",
       "                                             comment  \\\n",
       "0           The snow will last as long as my sadness   \n",
       "1                   I love experiencing new cultures   \n",
       "2  Look at that smooth easy catch of the ball. li...   \n",
       "3           I wish I could just run down this shore!   \n",
       "4  Psh, I've seen way better floral arrangements....   \n",
       "\n",
       "                         image_hash  \n",
       "0  1e22a9cf867d718551386b427c3b6d18  \n",
       "1     96472caea58db27769f1c282e2ac0  \n",
       "2    f09d8fb76822158de129acb0fef463  \n",
       "3    6e4ccc739ff44ed11da20ad9892317  \n",
       "4  e7e1844aa9e67cddc6ffe8804d76e45b  "
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['personality'] = df['personality'].astype('category').cat.codes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "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>personality</th>\n",
       "      <th>comment</th>\n",
       "      <th>image_hash</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>120</td>\n",
       "      <td>The snow will last as long as my sadness</td>\n",
       "      <td>1e22a9cf867d718551386b427c3b6d18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>I love experiencing new cultures</td>\n",
       "      <td>96472caea58db27769f1c282e2ac0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>130</td>\n",
       "      <td>Look at that smooth easy catch of the ball. li...</td>\n",
       "      <td>f09d8fb76822158de129acb0fef463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>216</td>\n",
       "      <td>I wish I could just run down this shore!</td>\n",
       "      <td>6e4ccc739ff44ed11da20ad9892317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>138</td>\n",
       "      <td>Psh, I've seen way better floral arrangements....</td>\n",
       "      <td>e7e1844aa9e67cddc6ffe8804d76e45b</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   personality                                            comment  \\\n",
       "0          120           The snow will last as long as my sadness   \n",
       "1            2                   I love experiencing new cultures   \n",
       "2          130  Look at that smooth easy catch of the ball. li...   \n",
       "3          216           I wish I could just run down this shore!   \n",
       "4          138  Psh, I've seen way better floral arrangements....   \n",
       "\n",
       "                         image_hash  \n",
       "0  1e22a9cf867d718551386b427c3b6d18  \n",
       "1     96472caea58db27769f1c282e2ac0  \n",
       "2    f09d8fb76822158de129acb0fef463  \n",
       "3    6e4ccc739ff44ed11da20ad9892317  \n",
       "4  e7e1844aa9e67cddc6ffe8804d76e45b  "
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = spacy.load('en_core_web_md')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/georgy/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:4: RuntimeWarning: invalid value encountered in true_divide\n",
      "  after removing the cwd from sys.path.\n"
     ]
    }
   ],
   "source": [
    "data=[]\n",
    "for i in range(len(df)):\n",
    "    sent2vec=nlp(df.comment[i]).vector\n",
    "    el=list(sent2vec/np.sqrt(np.sum(sent2vec**2)))\n",
    "    data.append(el) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def classify(test_sentence, persona_dic, data):\n",
    "    mydoc2vec=nlp(test_sentence).vector\n",
    "    mydoc2vec/=np.sqrt(np.sum(mydoc2vec**2))\n",
    "    N =CosineDistance().distance(data,mydoc2vec)\n",
    "    A=np.argsort(N)\n",
    "    return persona_dic[df.personality[A[0]]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "# persona=classify('Look at that smooth easy catch of the ball', persona_dic, data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "# persona"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "def engaging_image_captions(folder_path, personality):\n",
    "    out = {}\n",
    "    \n",
    "    reply = {}\n",
    "    reply['text']=personality\n",
    "    \n",
    "    f = []\n",
    "    for (dirpath, dirnames, filenames) in walk(folder_path):\n",
    "        f.extend(filenames)\n",
    "        break\n",
    "\n",
    "    for i in range(len(f)):\n",
    "        try:   \n",
    "            f[i]=folder_path +f[i]\n",
    "            with open(f[i], mode='rb') as file:\n",
    "                img = file.read()\n",
    "            encoded = base64.encodebytes(img).decode(\"utf-8\") \n",
    "\n",
    "            image = IM.open(io.BytesIO(b64decode(encoded))).convert('RGB')\n",
    "            reply['image'] = SHARED['image_loader'].extract(image)\n",
    "            SHARED['agent'].observe(reply)\n",
    "            model_res = SHARED['agent'].act()  \n",
    "            out[f[i]]=model_res['text']\n",
    "        except:\n",
    "            continue\n",
    "            \n",
    "    return out  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# captions_dic=engaging_image_captions(photo_folder_path, persona)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "# captions_dic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "def similarity(captions_dic, what_are_you_thinking_about):\n",
    "    out_={}\n",
    "    max_sim=float('inf')    \n",
    "    \n",
    "    \n",
    "    waryta2vec=nlp(what_are_you_thinking_about).vector\n",
    "    waryta2vec/=np.sqrt(np.sum(waryta2vec**2))\n",
    "    \n",
    "    \n",
    "    df= pd.DataFrame(list(captions_dic.items()), columns=['path', 'comment'])\n",
    "\n",
    "    data=[]\n",
    "    for i in range(len(df.comment)):\n",
    "        sent2vec=nlp(df.comment[i]).vector\n",
    "        el=list(sent2vec/np.sqrt(np.sum(sent2vec**2)))\n",
    "        data.append(el) \n",
    "        \n",
    "    N =CosineDistance().distance(data,waryta2vec)\n",
    "    A=np.argsort(N) \n",
    "       \n",
    "    return  df.path[A[0]], df.comment[A[0]]  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [],
   "source": [
    "# output_pic, output_phrase = similarity(captions_dic, what_are_you_thinking_about)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# USER INPUT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "what_are_you_thinking_about = \"I just love life\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [],
   "source": [
    "photo_folder_path = './photos/'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "persona=classify(what_are_you_thinking_about, persona_dic, data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [],
   "source": [
    "captions_dic=engaging_image_captions(photo_folder_path, persona)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_pic, output_phrase = similarity(captions_dic, what_are_you_thinking_about)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# OUTPUT TO POST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename=output_pic) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'I am so happy there are trees in the world to help us breathe.'"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "output_phrase"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}