approach_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import required libs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import rdflib\n",
    "from transformers import AutoTokenizer, AutoModelForSequenceClassification, T5ForConditionalGeneration, T5Tokenizer, pipeline\n",
    "from transformers.utils import logging\n",
    "from sentence_transformers import SentenceTransformer\n",
    "import torch\n",
    "import pandas as pd\n",
    "\n",
    "import os\n",
    "from pprint import pprint\n",
    "from IPython.display import display, Markdown\n",
    "\n",
    "from utils import triple_sentiment_analysis, test_entailment\n",
    "from utils.sparql_queries import find_all_triples_q\n",
    "from anchor_points_extractor import anchor_points_extractor\n",
    "from graph_explorator import graph_explorator\n",
    "from g2t_generator import g2t_generator\n",
    "\n",
    "os.environ[\"TOKENIZERS_PARALLELISM\"] = \"false\"\n",
    "\n",
    "device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu')\n",
    "logging.set_verbosity_error()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Set the current goal and the triples to avoid as anchor points\n",
    "\n",
    "For example, if you want to run the approach for the subgoal &ldquo;Use flood warning system to predict flood using weather forecastings&rdquo;, which has been created using the triples `Flood warning system Predicts Flood` and `Flood warning system Analyzes Forecasting`, you should use the following parameters:\n",
    "\n",
    "```python\n",
    "goal = \"Use flood warning system to predict flood using weather forecastings\"\n",
    "filtered_out_triples = [\"Flood warning system Predicts Flood\", \"Flood warning system Analyzes Forecasting\"]\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "goal = \"Anticipate the impact of floods on people\"\n",
    "filtered_out_triples = []"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import the example knowledge graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "domain_graph = rdflib.Graph()\n",
    "domain_graph.parse(\"./flooding_graph.rdf\");\n",
    "\n",
    "query_results = domain_graph.query(find_all_triples_q)\n",
    "triples = [list(map(str, [row[\"subject\"], row[\"predicate\"], row[\"object\"]])) for row in query_results.bindings]\n",
    "\n",
    "data = []\n",
    "for t in triples:\n",
    "    # simple triple\n",
    "    triple = \" \".join(t)\n",
    "    # triples serialized\n",
    "    triple_with_separator = [t]\n",
    "    list_goal_triples = [(triple, goal, triple_with_separator)]\n",
    "\n",
    "    for element in list_goal_triples:\n",
    "        row = {'TRIPLE': element[0], 'GOAL': element[1], 'TRIPLE_SERIALIZED': element[2]}\n",
    "        data.append(row)\n",
    "\n",
    "goal_triples_df = pd.DataFrame(data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import the models used"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_sts = SentenceTransformer('all-mpnet-base-v2')\n",
    "\n",
    "model_nli_name = \"ynie/roberta-large-snli_mnli_fever_anli_R1_R2_R3-nli\"\n",
    "tokenizer_nli = AutoTokenizer.from_pretrained(model_nli_name)\n",
    "model_nli = AutoModelForSequenceClassification.from_pretrained(model_nli_name).to(device)\n",
    "\n",
    "model_g2t = T5ForConditionalGeneration.from_pretrained(\"Inria-CEDAR/WebNLG20T5B\").to(device)\n",
    "tokenizer_g2t = T5Tokenizer.from_pretrained(\"t5-base\", model_max_length=512)\n",
    "\n",
    "sentiment_model_path = \"cardiffnlp/twitter-roberta-base-sentiment-latest\"\n",
    "sentiment_task = pipeline(\"sentiment-analysis\", model=sentiment_model_path, tokenizer=sentiment_model_path, device=device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract anchor points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>TRIPLE</th>\n",
       "      <th>GOAL</th>\n",
       "      <th>TRIPLE_SERIALIZED</th>\n",
       "      <th>SCORE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>53</th>\n",
       "      <td>Flood Causes Health and safety risks</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Health and safety risks]]</td>\n",
       "      <td>0.716851</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>Flood Causes Loss of life</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Loss of life]]</td>\n",
       "      <td>0.674230</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>Flood Causes Environmental damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Environmental damage]]</td>\n",
       "      <td>0.647110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>Flood warning system Analyzes Forecasting</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Analyzes, Forecasting]]</td>\n",
       "      <td>0.641735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>Flood Causes Business disruption</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Business disruption]]</td>\n",
       "      <td>0.629650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>Flood Causes Property damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Property damage]]</td>\n",
       "      <td>0.619106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>Flood Causes Evacuation of residents</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Evacuation of residents]]</td>\n",
       "      <td>0.615268</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Flood warning system Predicts Flood</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Predicts, Flood]]</td>\n",
       "      <td>0.610021</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                       TRIPLE  \\\n",
       "53       Flood Causes Health and safety risks   \n",
       "27                  Flood Causes Loss of life   \n",
       "14          Flood Causes Environmental damage   \n",
       "19  Flood warning system Analyzes Forecasting   \n",
       "31           Flood Causes Business disruption   \n",
       "35               Flood Causes Property damage   \n",
       "52       Flood Causes Evacuation of residents   \n",
       "5         Flood warning system Predicts Flood   \n",
       "\n",
       "                                         GOAL  \\\n",
       "53  Anticipate the impact of floods on people   \n",
       "27  Anticipate the impact of floods on people   \n",
       "14  Anticipate the impact of floods on people   \n",
       "19  Anticipate the impact of floods on people   \n",
       "31  Anticipate the impact of floods on people   \n",
       "35  Anticipate the impact of floods on people   \n",
       "52  Anticipate the impact of floods on people   \n",
       "5   Anticipate the impact of floods on people   \n",
       "\n",
       "                                  TRIPLE_SERIALIZED     SCORE  \n",
       "53       [[Flood, Causes, Health and safety risks]]  0.716851  \n",
       "27                  [[Flood, Causes, Loss of life]]  0.674230  \n",
       "14          [[Flood, Causes, Environmental damage]]  0.647110  \n",
       "19  [[Flood warning system, Analyzes, Forecasting]]  0.641735  \n",
       "31           [[Flood, Causes, Business disruption]]  0.629650  \n",
       "35               [[Flood, Causes, Property damage]]  0.619106  \n",
       "52       [[Flood, Causes, Evacuation of residents]]  0.615268  \n",
       "5         [[Flood warning system, Predicts, Flood]]  0.610021  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "anchor_points_df = anchor_points_extractor(goal_triples_df, model_sts, filtered_out_triples).copy()\n",
    "anchor_points_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Transform negative triples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>PREMISE</th>\n",
       "      <th>HYPOTHESIS</th>\n",
       "      <th>PREMISE_SERIALIZED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Prevent that Flood Causes Health and safety risks</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Health and safety risks]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Prevent that Flood Causes Loss of life</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Loss of life]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Prevent that Flood Causes Environmental damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Environmental damage]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Flood warning system Analyzes Forecasting</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Analyzes, Forecasting]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Prevent that Flood Causes Business disruption</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Business disruption]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Prevent that Flood Causes Property damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Property damage]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Prevent that Flood Causes Evacuation of residents</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Evacuation of residents]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Flood warning system Predicts Flood</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Predicts, Flood]]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             PREMISE  \\\n",
       "0  Prevent that Flood Causes Health and safety risks   \n",
       "1             Prevent that Flood Causes Loss of life   \n",
       "2     Prevent that Flood Causes Environmental damage   \n",
       "3          Flood warning system Analyzes Forecasting   \n",
       "4      Prevent that Flood Causes Business disruption   \n",
       "5          Prevent that Flood Causes Property damage   \n",
       "6  Prevent that Flood Causes Evacuation of residents   \n",
       "7                Flood warning system Predicts Flood   \n",
       "\n",
       "                                  HYPOTHESIS  \\\n",
       "0  Anticipate the impact of floods on people   \n",
       "1  Anticipate the impact of floods on people   \n",
       "2  Anticipate the impact of floods on people   \n",
       "3  Anticipate the impact of floods on people   \n",
       "4  Anticipate the impact of floods on people   \n",
       "5  Anticipate the impact of floods on people   \n",
       "6  Anticipate the impact of floods on people   \n",
       "7  Anticipate the impact of floods on people   \n",
       "\n",
       "                                PREMISE_SERIALIZED  \n",
       "0       [[Flood, Causes, Health and safety risks]]  \n",
       "1                  [[Flood, Causes, Loss of life]]  \n",
       "2          [[Flood, Causes, Environmental damage]]  \n",
       "3  [[Flood warning system, Analyzes, Forecasting]]  \n",
       "4           [[Flood, Causes, Business disruption]]  \n",
       "5               [[Flood, Causes, Property damage]]  \n",
       "6       [[Flood, Causes, Evacuation of residents]]  \n",
       "7        [[Flood warning system, Predicts, Flood]]  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "anchor_points_df[\"SENTIMENT\"] = anchor_points_df[\"TRIPLE_SERIALIZED\"].apply(\n",
    "        lambda triple: triple_sentiment_analysis(triple[0], sentiment_task, neutral_predicates=[\"is a type of\"])[0])\n",
    "anchor_points_df.rename(columns={'TRIPLE': 'PREMISE', 'GOAL': 'HYPOTHESIS'}, inplace=True)\n",
    "transformed_triples_premise = []\n",
    "\n",
    "for triple, sentiment in zip(anchor_points_df[\"PREMISE\"], anchor_points_df[\"SENTIMENT\"]):\n",
    "    if sentiment == \"negative\":\n",
    "        ### Transformation\n",
    "        transformed_triples_premise.append(\"Prevent that \" + triple)\n",
    "    else:\n",
    "        transformed_triples_premise.append(triple)\n",
    "transformed_anchor_points = pd.DataFrame(transformed_triples_premise, columns=[\"PREMISE\"])\n",
    "transformed_anchor_points[\"HYPOTHESIS\"] = anchor_points_df[\"HYPOTHESIS\"].values\n",
    "transformed_anchor_points[\"PREMISE_SERIALIZED\"] = anchor_points_df[\"TRIPLE_SERIALIZED\"].values\n",
    "\n",
    "transformed_anchor_points"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Verify entailment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>PREMISE</th>\n",
       "      <th>HYPOTHESIS</th>\n",
       "      <th>PREMISE_SERIALIZED</th>\n",
       "      <th>ENTAILMENT</th>\n",
       "      <th>NEUTRAL</th>\n",
       "      <th>CONTRADICTION</th>\n",
       "      <th>NLI_LABEL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Prevent that Flood Causes Evacuation of residents</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Evacuation of residents]]</td>\n",
       "      <td>0.86817</td>\n",
       "      <td>0.11147</td>\n",
       "      <td>0.02036</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Prevent that Flood Causes Loss of life</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Loss of life]]</td>\n",
       "      <td>0.78269</td>\n",
       "      <td>0.17343</td>\n",
       "      <td>0.04388</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Prevent that Flood Causes Health and safety risks</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Health and safety risks]]</td>\n",
       "      <td>0.70614</td>\n",
       "      <td>0.21396</td>\n",
       "      <td>0.07990</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Prevent that Flood Causes Business disruption</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Business disruption]]</td>\n",
       "      <td>0.41095</td>\n",
       "      <td>0.35846</td>\n",
       "      <td>0.23059</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Flood warning system Predicts Flood</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Predicts, Flood]]</td>\n",
       "      <td>0.38620</td>\n",
       "      <td>0.59991</td>\n",
       "      <td>0.01389</td>\n",
       "      <td>NEUTRAL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Prevent that Flood Causes Property damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Property damage]]</td>\n",
       "      <td>0.35060</td>\n",
       "      <td>0.34778</td>\n",
       "      <td>0.30162</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Flood warning system Analyzes Forecasting</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood warning system, Analyzes, Forecasting]]</td>\n",
       "      <td>0.21787</td>\n",
       "      <td>0.76808</td>\n",
       "      <td>0.01405</td>\n",
       "      <td>NEUTRAL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Prevent that Flood Causes Environmental damage</td>\n",
       "      <td>Anticipate the impact of floods on people</td>\n",
       "      <td>[[Flood, Causes, Environmental damage]]</td>\n",
       "      <td>0.16007</td>\n",
       "      <td>0.38834</td>\n",
       "      <td>0.45159</td>\n",
       "      <td>CONTRADICTION</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             PREMISE  \\\n",
       "6  Prevent that Flood Causes Evacuation of residents   \n",
       "1             Prevent that Flood Causes Loss of life   \n",
       "0  Prevent that Flood Causes Health and safety risks   \n",
       "4      Prevent that Flood Causes Business disruption   \n",
       "7                Flood warning system Predicts Flood   \n",
       "5          Prevent that Flood Causes Property damage   \n",
       "3          Flood warning system Analyzes Forecasting   \n",
       "2     Prevent that Flood Causes Environmental damage   \n",
       "\n",
       "                                  HYPOTHESIS  \\\n",
       "6  Anticipate the impact of floods on people   \n",
       "1  Anticipate the impact of floods on people   \n",
       "0  Anticipate the impact of floods on people   \n",
       "4  Anticipate the impact of floods on people   \n",
       "7  Anticipate the impact of floods on people   \n",
       "5  Anticipate the impact of floods on people   \n",
       "3  Anticipate the impact of floods on people   \n",
       "2  Anticipate the impact of floods on people   \n",
       "\n",
       "                                PREMISE_SERIALIZED  ENTAILMENT  NEUTRAL  \\\n",
       "6       [[Flood, Causes, Evacuation of residents]]     0.86817  0.11147   \n",
       "1                  [[Flood, Causes, Loss of life]]     0.78269  0.17343   \n",
       "0       [[Flood, Causes, Health and safety risks]]     0.70614  0.21396   \n",
       "4           [[Flood, Causes, Business disruption]]     0.41095  0.35846   \n",
       "7        [[Flood warning system, Predicts, Flood]]     0.38620  0.59991   \n",
       "5               [[Flood, Causes, Property damage]]     0.35060  0.34778   \n",
       "3  [[Flood warning system, Analyzes, Forecasting]]     0.21787  0.76808   \n",
       "2          [[Flood, Causes, Environmental damage]]     0.16007  0.38834   \n",
       "\n",
       "   CONTRADICTION      NLI_LABEL  \n",
       "6        0.02036     ENTAILMENT  \n",
       "1        0.04388     ENTAILMENT  \n",
       "0        0.07990     ENTAILMENT  \n",
       "4        0.23059     ENTAILMENT  \n",
       "7        0.01389        NEUTRAL  \n",
       "5        0.30162     ENTAILMENT  \n",
       "3        0.01405        NEUTRAL  \n",
       "2        0.45159  CONTRADICTION  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "entailment_result = test_entailment(transformed_anchor_points, tokenizer_nli, model_nli)\n",
    "entailment_result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Explore graph to improve contextualization\n",
    "\n",
    "Add neighbor triples to anchor points to further contextualize them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>SUBGOALS</th>\n",
       "      <th>SUBGOALS_SERIALIZED</th>\n",
       "      <th>SCORE</th>\n",
       "      <th>NLI_LABEL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Prevent that Flood Causes Evacuation of residents</td>\n",
       "      <td>[[Flood, Causes, Evacuation of residents]]</td>\n",
       "      <td>0.86817</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Prevent that Flood Causes Loss of life</td>\n",
       "      <td>[[Flood, Causes, Loss of life]]</td>\n",
       "      <td>0.78269</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Prevent that Flood Causes Health and safety risks</td>\n",
       "      <td>[[Flood, Causes, Health and safety risks]]</td>\n",
       "      <td>0.70614</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Prevent that Flood Causes Business disruption</td>\n",
       "      <td>[[Flood, Causes, Business disruption]]</td>\n",
       "      <td>0.41095</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Flood Causes Loss of life. Flood warning syste...</td>\n",
       "      <td>[[Flood, Causes, Loss of life], [Flood warning...</td>\n",
       "      <td>0.82918</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Prevent that Flood Causes Property damage</td>\n",
       "      <td>[[Flood, Causes, Property damage]]</td>\n",
       "      <td>0.35060</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Flood warning system Recommends Evacuation of ...</td>\n",
       "      <td>[[Flood warning system, Recommends, Evacuation...</td>\n",
       "      <td>0.65599</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Flood Causes Evacuation of residents. Prevent ...</td>\n",
       "      <td>[[Flood, Causes, Evacuation of residents], [Fl...</td>\n",
       "      <td>0.74740</td>\n",
       "      <td>ENTAILMENT</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            SUBGOALS  \\\n",
       "0  Prevent that Flood Causes Evacuation of residents   \n",
       "1             Prevent that Flood Causes Loss of life   \n",
       "2  Prevent that Flood Causes Health and safety risks   \n",
       "3      Prevent that Flood Causes Business disruption   \n",
       "4  Flood Causes Loss of life. Flood warning syste...   \n",
       "5          Prevent that Flood Causes Property damage   \n",
       "2  Flood warning system Recommends Evacuation of ...   \n",
       "4  Flood Causes Evacuation of residents. Prevent ...   \n",
       "\n",
       "                                 SUBGOALS_SERIALIZED    SCORE   NLI_LABEL  \n",
       "0         [[Flood, Causes, Evacuation of residents]]  0.86817  ENTAILMENT  \n",
       "1                    [[Flood, Causes, Loss of life]]  0.78269  ENTAILMENT  \n",
       "2         [[Flood, Causes, Health and safety risks]]  0.70614  ENTAILMENT  \n",
       "3             [[Flood, Causes, Business disruption]]  0.41095  ENTAILMENT  \n",
       "4  [[Flood, Causes, Loss of life], [Flood warning...  0.82918  ENTAILMENT  \n",
       "5                 [[Flood, Causes, Property damage]]  0.35060  ENTAILMENT  \n",
       "2  [[Flood warning system, Recommends, Evacuation...  0.65599  ENTAILMENT  \n",
       "4  [[Flood, Causes, Evacuation of residents], [Fl...  0.74740  ENTAILMENT  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "entailed_triples_df = graph_explorator(entailment_result, goal, domain_graph, tokenizer_nli, model_nli)\n",
    "entailed_triples_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generate text from the triples identified as relevant"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "#### Unique triples"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['Flood', 'Causes', 'Evacuation of residents'],\n",
      " ['Flood', 'Causes', 'Loss of life'],\n",
      " ['Flood', 'Causes', 'Health and safety risks'],\n",
      " ['Flood', 'Causes', 'Business disruption'],\n",
      " ['Flood warning system', 'Predicts', 'Flood'],\n",
      " ['Flood', 'Causes', 'Property damage'],\n",
      " ['Flood warning system', 'Recommends', 'Evacuation of residents'],\n",
      " ['Flood warning system', 'Analyzes', 'Forecasting'],\n",
      " ['Flood', 'Causes', 'Environmental damage']]\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "---------------------------"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/markdown": [
       "#### Generated texts"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[AVOID] evacuation of residents is one of the causes of the flood\n",
      "[AVOID] floods can cause loss of life\n",
      "[AVOID] floods can cause health and safety risks\n",
      "[AVOID] floods can cause business disruption\n",
      "[ACHIEVE] the flood warning system predicts a flood\n",
      "[AVOID] flooding can cause property damage\n",
      "[ACHIEVE] the flood warning system recommends evacuation of residents and analyzes forecasting\n",
      "[AVOID] floods can cause environmental damage\n"
     ]
    }
   ],
   "source": [
    "if not entailed_triples_df.empty:\n",
    "    all_triples_entailed = [triple for triples in entailed_triples_df[\"SUBGOALS_SERIALIZED\"].tolist() for triple in triples]\n",
    "\n",
    "    if not entailed_triples_df.empty:\n",
    "        all_triples_entailed.append(triples[0] for triples in entailed_triples_df[\"SUBGOALS_SERIALIZED\"].tolist())\n",
    "\n",
    "    unique_triples_entailed = []\n",
    "    for triple in all_triples_entailed:\n",
    "        if (triple not in unique_triples_entailed) and (type(triple) is list):\n",
    "            unique_triples_entailed.append(triple)\n",
    "\n",
    "    display(Markdown(\"#### Unique triples\"))\n",
    "    pprint(unique_triples_entailed)\n",
    "    display(Markdown(\"---------------------------\"))\n",
    "\n",
    "    triples_already_processed = []\n",
    "\n",
    "    display(Markdown(\"#### Generated texts\"))\n",
    "\n",
    "    for idx, row in entailed_triples_df.iterrows():\n",
    "        triples_to_process = []\n",
    "        for triple in row[\"SUBGOALS_SERIALIZED\"]:\n",
    "            if (triple not in triples_already_processed) and (type(triple) is list):\n",
    "                triples_to_process.append(triple)\n",
    "        if triples_to_process:\n",
    "            prediction = g2t_generator(triples_to_process, model=model_g2t, tokenizer=tokenizer_g2t)[0]\n",
    "            text_version = row[\"SUBGOALS\"].split(\". \")[-1]\n",
    "            if \"Prevent that\" in text_version:\n",
    "                prediction = \"[AVOID] \" + prediction\n",
    "            else:\n",
    "                prediction = \"[ACHIEVE] \" + prediction\n",
    "            print(prediction)\n",
    "\n",
    "        triples_already_processed.extend(triples_to_process)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}