{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import pickle\n",
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "SCIDMT_PATH = {\n",
    "    'DICT': './SciDMT_dict.json',\n",
    "\n",
    "    # machine learning inputs at sentence level\n",
    "    'sent_xy': './SciDMT_sentences.p', \n",
    "    'sent_eval': './SciDMT_E_sentences.json',\n",
    "    'sent_split': './SciDMT_sentences_split.json',\n",
    "\n",
    "    # document level inputs\n",
    "    'doc_split': './SciDMT_split.json',\n",
    "    'doc_eval': './SciDMT_E_human_annotations.json',\n",
    "    'doc_text_and_meta': './SciDMT_papers.csv',\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_dict_structure(d, indent=0, indent_str='   '):\n",
    "    for key, value in d.items():\n",
    "        print(indent_str * indent + ' ' +str(key))\n",
    "        if isinstance(value, dict):\n",
    "            print_dict_structure(value, indent+1)\n",
    "        else:\n",
    "            if type(value) == list:\n",
    "                des = f'| len={len(value)} | first 3 entries={value[:3]}'\n",
    "            else:\n",
    "                des = ''\n",
    "            print(indent_str * (indent+1) + ' ' + str(type(value)) + des)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['datasets', 'methods', 'tasks'])\n"
     ]
    }
   ],
   "source": [
    "# Load SciDMT dictionary for entities\n",
    "DICT = json.load(open(SCIDMT_PATH['DICT'], 'r'))\n",
    "print(DICT.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load SciDMT evaluation set in sentence level\n",
    "scidmt_e = pd.read_json(SCIDMT_PATH['sent_eval'])\n",
    "\n",
    "# load x_test, y_test\n",
    "X_test, y_test = scidmt_e['X'].to_list(), scidmt_e['y'].to_list()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " is_contain\n",
      "    datasets\n",
      "       <class 'list'>| len=48049 | first 3 entries=[True, False, True]\n",
      "    tasks\n",
      "       <class 'list'>| len=48049 | first 3 entries=[True, True, True]\n",
      "    methods\n",
      "       <class 'list'>| len=48049 | first 3 entries=[True, True, True]\n",
      "    all\n",
      "       <class 'list'>| len=48049 | first 3 entries=[True, True, True]\n",
      " is_test\n",
      "    datasets\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      "    tasks\n",
      "       <class 'list'>| len=48049 | first 3 entries=[True, False, False]\n",
      "    methods\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, True, False]\n",
      "    all\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      " doc_pids\n",
      "    <class 'list'>| len=48049 | first 3 entries=[51881821, 51881855, 51881912]\n",
      " is_0shot\n",
      "    datasets\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      "    methods\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      "    tasks\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      "    all\n",
      "       <class 'list'>| len=48049 | first 3 entries=[False, False, False]\n",
      " SHOT0\n",
      "    datasets\n",
      "       entity_indexs\n",
      "          <class 'list'>| len=10 | first 3 entries=[1561, 2889, 2810]\n",
      "       ml_inputs_idxs\n",
      "          <class 'list'>| len=147 | first 3 entries=[461223, 461224, 461225]\n",
      "    tasks\n",
      "       entity_indexs\n",
      "          <class 'list'>| len=10 | first 3 entries=[766, 1487, 1548]\n",
      "       ml_inputs_idxs\n",
      "          <class 'list'>| len=305 | first 3 entries=[646887, 646888, 646889]\n",
      "    methods\n",
      "       entity_indexs\n",
      "          <class 'list'>| len=10 | first 3 entries=[605, 1324, 1099]\n",
      "       ml_inputs_idxs\n",
      "          <class 'list'>| len=128 | first 3 entries=[1154530, 1154548, 1154550]\n"
     ]
    }
   ],
   "source": [
    "# load document level split\n",
    "DOC_SPLIT = json.load(open(SCIDMT_PATH['doc_split'], 'r'))\n",
    "print_dict_structure(DOC_SPLIT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " is_contain\n",
      "    datasets\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[True, False, False]\n",
      "    tasks\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[False, True, True]\n",
      "    methods\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[False, False, False]\n",
      "    all\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[True, True, True]\n",
      " is_test\n",
      "    datasets\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[False, False, False]\n",
      "    tasks\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[True, True, True]\n",
      "    methods\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[True, True, True]\n",
      "    all\n",
      "       <class 'list'>| len=1128148 | first 3 entries=[False, False, False]\n"
     ]
    }
   ],
   "source": [
    "# load sentence level train/test split\n",
    "with open(SCIDMT_PATH['sent_split'], 'r') as f:\n",
    "    SPLIT = json.load(f)\n",
    "print_dict_structure(SPLIT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_1610900/687978490.py:4: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  X_train = np.array(X)[is_train]#[:5000] ##debug\n",
      "/tmp/ipykernel_1610900/687978490.py:5: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  y_train = np.array(y)[is_train]#[:5000] ##debug\n"
     ]
    }
   ],
   "source": [
    "# load x_train, y_train\n",
    "# is_test = true are not all human annotated. Thus, we have a seperate test file\n",
    "is_train = np.logical_and(SPLIT['is_test']['all']==False, SPLIT['is_contain']['all'] == True)\n",
    "X, y, _, _, _ = pickle.load(open(SCIDMT_PATH['sent_xy'], 'rb'))\n",
    "X_train = np.array(X)[is_train]#[:5000] ##debug\n",
    "y_train = np.array(y)[is_train]#[:5000] ##debug"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py39",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}