main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import operator\n",
    "import os\n",
    "from pathlib import Path\n",
    "from pprint import pprint\n",
    "import re\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load data and preprocess"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Map from test suite tag to high-level circuit.\n",
    "circuits = {\n",
    "    \"Licensing\": [\"npi\", \"reflexive\"],\n",
    "    \"Long-Distance Dependencies\": [\"fgd\", \"position\", \"embed\", \"number\", \"comparison\"],\n",
    "    \"Garden-Path Effects\": [\"gardenpath\", \"npz\", \"mvrr\"],\n",
    "    \"Gross Syntactic State\": [\"subordination\"],\n",
    "    \"Center Embedding\": [\"center\"],\n",
    "    \"Transformations\": [\"passive\", \"cleft\"],\n",
    "}\n",
    "\n",
    "tag_to_circuit = {tag: circuit\n",
    "                  for circuit, tags in circuits.items()\n",
    "                  for tag in tags}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Exclusions\n",
    "exclude_suite_re = re.compile(r\"^cleft_modifier|^fgd-embed[34]|^npz_(ambig)\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.chdir(\"..\")\n",
    "ppl_data_path = Path(\"data/raw/perplexity.csv\")\n",
    "test_suite_results_path = Path(\"data/raw/test_suite_results.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dropping 1686 results / 3 suites due to exclusions.\n",
      "Tags missing circuit:  nn\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:11: UserWarning: This pattern has match groups. To actually get the groups, use str.extract.\n",
      "  # This is added back by InteractiveShellApp.init_path()\n"
     ]
    }
   ],
   "source": [
    "perplexity_df = pd.read_csv(ppl_data_path, index_col=[\"model\", \"corpus\", \"seed\"])\n",
    "perplexity_df.index.set_names(\"model_name\", level=0, inplace=True)\n",
    "results_df = pd.read_csv(test_suite_results_path, index_col=0)\n",
    "\n",
    "# Split model_id into constituent parts\n",
    "model_ids = results_df.model.str.split(\"_\", expand=True).rename(columns={0: \"model_name\", 1: \"corpus\", 2: \"seed\"})\n",
    "results_df = results_df.join(model_ids).drop(columns=[\"model\"])\n",
    "results_df[\"seed\"] = results_df.seed.astype(int)\n",
    "\n",
    "# Exclude test suites\n",
    "exclude_filter = results_df.suite.str.contains(exclude_suite_re)\n",
    "print(\"Dropping %i results / %i suites due to exclusions.\"\n",
    "      % (exclude_filter.sum(), len(results_df[exclude_filter].suite.unique())))\n",
    "results_df = results_df[~exclude_filter]\n",
    "\n",
    "# Add tags\n",
    "results_df[\"tag\"] = results_df.suite.transform(lambda s: re.split(r\"[-_0-9]\", s)[0])\n",
    "results_df[\"circuit\"] = results_df.tag.map(tag_to_circuit)\n",
    "tags_missing_circuit = set(results_df.tag.unique()) - set(tag_to_circuit.keys())\n",
    "if tags_missing_circuit:\n",
    "    print(\"Tags missing circuit: \", \", \".join(tags_missing_circuit))"
   ]
  },
  {
   "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>suite</th>\n",
       "      <th>item</th>\n",
       "      <th>correct</th>\n",
       "      <th>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>tag</th>\n",
       "      <th>circuit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>number_src</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-xs</td>\n",
       "      <td>922</td>\n",
       "      <td>number</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>number_src</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-sm</td>\n",
       "      <td>120</td>\n",
       "      <td>number</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>number_src</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-md</td>\n",
       "      <td>111</td>\n",
       "      <td>number</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>number_src</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>number</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>npi_orc_any</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-xs</td>\n",
       "      <td>922</td>\n",
       "      <td>npi</td>\n",
       "      <td>Licensing</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         suite  item  correct       model_name    corpus  seed     tag  \\\n",
       "0   number_src     0    False  ordered-neurons  bllip-xs   922  number   \n",
       "0   number_src     0    False  ordered-neurons  bllip-sm   120  number   \n",
       "0   number_src     0    False  ordered-neurons  bllip-md   111  number   \n",
       "0   number_src     0    False  ordered-neurons  bllip-lg   111  number   \n",
       "0  npi_orc_any     0    False  ordered-neurons  bllip-xs   922     npi   \n",
       "\n",
       "                      circuit  \n",
       "0  Long-Distance Dependencies  \n",
       "0  Long-Distance Dependencies  \n",
       "0  Long-Distance Dependencies  \n",
       "0  Long-Distance Dependencies  \n",
       "0                   Licensing  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Checks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Each model--corpus--seed should have perplexity data.\n",
    "ids_from_results = results_df.set_index([\"model_name\", \"corpus\", \"seed\"]).sort_index().index\n",
    "ids_from_ppl = perplexity_df.sort_index().index\n",
    "diff = set(ids_from_results) - set(ids_from_ppl)\n",
    "if diff:\n",
    "    print(\"Missing perplexity results for:\")\n",
    "    pprint(diff)\n",
    "    raise ValueError(\"Each model--corpus--seed must have perplexity data.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Main analyses\n",
    "\n",
    "1. barplot ranking model accuracies\n",
    "2. scatter plot with ppl on x-axis and SG score on y-axis (for a given dataset size -- or maybe all of them together?)\n",
    "3. variance in ppl vs variance in SG score for a single model across seeds and/or sizes\n",
    "\n",
    "Test suite analyses\n",
    "\n",
    "1. within-tag/circuit ppl-SG correlations\n",
    "2. circuit-circuit coordination heatmap\n",
    "3. robustness to stability modification}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Global settings\n",
    "\n",
    "e.g. to maintain consistent hues across model graphs, etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_order = sorted(set(results_df.model_name))\n",
    "corpus_order = [\"bllip-lg\", \"bllip-md\", \"bllip-sm\", \"bllip-xs\"]\n",
    "circuit_order = sorted([c for c in results_df.circuit.dropna().unique()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data prep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "suites_df = results_df.groupby([\"model_name\", \"corpus\", \"seed\", \"suite\"]).correct.mean().reset_index()\n",
    "suites_df[\"tag\"] = suites_df.suite.transform(lambda s: re.split(r\"[-_0-9]\", s)[0])\n",
    "suites_df[\"circuit\"] = suites_df.tag.map(tag_to_circuit)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>correct</th>\n",
       "      <th>pid</th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_ppl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>0.361735</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-md</td>\n",
       "      <td>111</td>\n",
       "      <td>0.361735</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.02</td>\n",
       "      <td>55.96</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-sm</td>\n",
       "      <td>111</td>\n",
       "      <td>0.373540</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3.99</td>\n",
       "      <td>53.85</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-sm</td>\n",
       "      <td>120</td>\n",
       "      <td>0.359645</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3.99</td>\n",
       "      <td>54.06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-xs</td>\n",
       "      <td>111</td>\n",
       "      <td>0.381741</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.30</td>\n",
       "      <td>73.46</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        model_name    corpus  seed   correct  pid  test_loss  test_ppl\n",
       "0  ordered-neurons  bllip-lg   111  0.361735  NaN       4.03     56.38\n",
       "1  ordered-neurons  bllip-md   111  0.361735  NaN       4.02     55.96\n",
       "2  ordered-neurons  bllip-sm   111  0.373540  NaN       3.99     53.85\n",
       "3  ordered-neurons  bllip-sm   120  0.359645  NaN       3.99     54.06\n",
       "4  ordered-neurons  bllip-xs   111  0.381741  NaN       4.30     73.46"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Join PPL and accuracy data.\n",
    "joined_data = suites_df.groupby([\"model_name\", \"corpus\", \"seed\"]).correct.agg(\"mean\")\n",
    "joined_data = pd.DataFrame(joined_data).join(perplexity_df).reset_index()\n",
    "joined_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>circuit</th>\n",
       "      <th>correct</th>\n",
       "      <th>pid</th>\n",
       "      <th>test_loss</th>\n",
       "      <th>test_ppl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>Center Embedding</td>\n",
       "      <td>0.571429</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>0.419643</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>Licensing</td>\n",
       "      <td>0.097115</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>Long-Distance Dependencies</td>\n",
       "      <td>0.488210</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>Transformations</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4.03</td>\n",
       "      <td>56.38</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        model_name    corpus  seed                     circuit   correct  pid  \\\n",
       "0  ordered-neurons  bllip-lg   111            Center Embedding  0.571429  NaN   \n",
       "1  ordered-neurons  bllip-lg   111         Garden-Path Effects  0.419643  NaN   \n",
       "2  ordered-neurons  bllip-lg   111                   Licensing  0.097115  NaN   \n",
       "3  ordered-neurons  bllip-lg   111  Long-Distance Dependencies  0.488210  NaN   \n",
       "4  ordered-neurons  bllip-lg   111             Transformations  1.000000  NaN   \n",
       "\n",
       "   test_loss  test_ppl  \n",
       "0       4.03     56.38  \n",
       "1       4.03     56.38  \n",
       "2       4.03     56.38  \n",
       "3       4.03     56.38  \n",
       "4       4.03     56.38  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Join PPL and accuracy data, splitting on circuit.\n",
    "joined_data_circuits = suites_df.groupby([\"model_name\", \"corpus\", \"seed\", \"circuit\"]).correct.agg(\"mean\")\n",
    "joined_data_circuits = pd.DataFrame(joined_data_circuits).reset_index().set_index([\"model_name\", \"corpus\", \"seed\"]).join(perplexity_df).reset_index()\n",
    "joined_data_circuits.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:16: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  app.launch_new_instance()\n"
     ]
    },
    {
     "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>suite</th>\n",
       "      <th>item</th>\n",
       "      <th>correct</th>\n",
       "      <th>model_name</th>\n",
       "      <th>corpus</th>\n",
       "      <th>seed</th>\n",
       "      <th>tag</th>\n",
       "      <th>circuit</th>\n",
       "      <th>has_modifier</th>\n",
       "      <th>test_suite_base</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>133</th>\n",
       "      <td>gardenpath_mod</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-xs</td>\n",
       "      <td>922</td>\n",
       "      <td>gardenpath</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>True</td>\n",
       "      <td>gardenpath</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>133</th>\n",
       "      <td>gardenpath_mod</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-sm</td>\n",
       "      <td>111</td>\n",
       "      <td>gardenpath</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>True</td>\n",
       "      <td>gardenpath</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>133</th>\n",
       "      <td>gardenpath_mod</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-md</td>\n",
       "      <td>111</td>\n",
       "      <td>gardenpath</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>True</td>\n",
       "      <td>gardenpath</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>133</th>\n",
       "      <td>gardenpath_mod</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-lg</td>\n",
       "      <td>111</td>\n",
       "      <td>gardenpath</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>True</td>\n",
       "      <td>gardenpath</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>133</th>\n",
       "      <td>gardenpath_mod</td>\n",
       "      <td>0</td>\n",
       "      <td>False</td>\n",
       "      <td>ordered-neurons</td>\n",
       "      <td>bllip-xs</td>\n",
       "      <td>922</td>\n",
       "      <td>gardenpath</td>\n",
       "      <td>Garden-Path Effects</td>\n",
       "      <td>True</td>\n",
       "      <td>gardenpath</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              suite  item  correct       model_name    corpus  seed  \\\n",
       "133  gardenpath_mod     0    False  ordered-neurons  bllip-xs   922   \n",
       "133  gardenpath_mod     0    False  ordered-neurons  bllip-sm   111   \n",
       "133  gardenpath_mod     0    False  ordered-neurons  bllip-md   111   \n",
       "133  gardenpath_mod     0    False  ordered-neurons  bllip-lg   111   \n",
       "133  gardenpath_mod     0    False  ordered-neurons  bllip-xs   922   \n",
       "\n",
       "            tag              circuit has_modifier test_suite_base  \n",
       "133  gardenpath  Garden-Path Effects         True      gardenpath  \n",
       "133  gardenpath  Garden-Path Effects         True      gardenpath  \n",
       "133  gardenpath  Garden-Path Effects         True      gardenpath  \n",
       "133  gardenpath  Garden-Path Effects         True      gardenpath  \n",
       "133  gardenpath  Garden-Path Effects         True      gardenpath  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Analyze stability to modification.\n",
    "def has_modifier(ts):\n",
    "    if ts.endswith((\"_modifier\", \"_mod\")):\n",
    "        return True\n",
    "    else:\n",
    "        return None\n",
    "results_df[\"has_modifier\"] = results_df.suite.transform(has_modifier)\n",
    "\n",
    "# Mark \"non-modifier\" test suites\n",
    "modifier_ts = results_df[results_df.has_modifier == True].suite.unique()\n",
    "no_modifier_ts = [re.sub(r\"_mod(ifier)?$\", \"\", ts) for ts in modifier_ts]\n",
    "results_df.loc[results_df.suite.isin(no_modifier_ts), \"has_modifier\"] = False\n",
    "# Store subset of test suites which have definite modifier/no-modifier marking\n",
    "results_df_mod = results_df[~(results_df.has_modifier.isna())]\n",
    "# Get base test suite (without modifier/no-modifier marking)\n",
    "results_df_mod[\"test_suite_base\"] = results_df_mod.suite.transform(lambda ts: ts.strip(\"_no-modifier\").strip(\"_modifier\"))\n",
    "results_df_mod.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Accuracy across models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'Accuracy')"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.barplot(data=results_df.reset_index(), x=\"model_name\", y=\"correct\")\n",
    "\n",
    "plt.xlabel(\"Model\")\n",
    "plt.ylabel(\"Accuracy\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Accuracy vs perplexity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus_to_size = {\n",
    "    \"bllip-xs\": 2,\n",
    "    \"bllip-sm\": 8,\n",
    "    \"bllip-md\": 14,\n",
    "    \"bllip-lg\": 20,\n",
    "}\n",
    "\n",
    "model_colors = dict(zip(model_order,\n",
    "                        ['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728', '#9467bd', '#8c564b', '#e377c2', '#7f7f7f', '#bcbd22', '#17becf']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x2b1ae3ddb630>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, ax = plt.subplots(figsize=(10, 10))\n",
    "# sns.regplot(data=graph_data, x=\"test_ppl\", y=\"correct\",\n",
    "#             scatter_kws={\"s\": graph_data[\"corpus\"].map(corpus_to_size),\n",
    "#                          \"facecolors\": graph_data.model.map(model_colors)})\n",
    "sns.scatterplot(data=joined_data, x=\"test_ppl\", y=\"correct\",\n",
    "                hue=\"model_name\", size=\"corpus\",\n",
    "                hue_order=model_order, size_order=corpus_order)\n",
    "plt.xlabel(\"Test corpus perplexity\")\n",
    "plt.ylabel(\"SyntaxGym score\")\n",
    "plt.legend(bbox_to_anchor=(1.04,1), loc=\"upper left\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x2b1ae3e0b8d0>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 436.75x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.lmplot(data=joined_data, x=\"test_ppl\", y=\"correct\",\n",
    "           hue=\"corpus\", truncate=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x2b1ae3e0bc88>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x216 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "g = sns.FacetGrid(data=joined_data_circuits, col=\"circuit\")\n",
    "g.map(sns.scatterplot, \"test_ppl\", \"correct\", \"model_name\",\n",
    "      hue_order=model_order)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "model_name       corpus    circuit                   \n",
       "ordered-neurons  bllip-lg  Center Embedding              0.571429\n",
       "                           Garden-Path Effects           0.419643\n",
       "                           Licensing                     0.097115\n",
       "                           Long-Distance Dependencies    0.488210\n",
       "                           Transformations               1.000000\n",
       "                 bllip-md  Center Embedding              0.571429\n",
       "                           Garden-Path Effects           0.419643\n",
       "                           Licensing                     0.097115\n",
       "                           Long-Distance Dependencies    0.488210\n",
       "                           Transformations               1.000000\n",
       "                 bllip-sm  Center Embedding              0.583257\n",
       "                           Garden-Path Effects           0.413112\n",
       "                           Licensing                     0.127452\n",
       "                           Long-Distance Dependencies    0.459234\n",
       "                           Transformations               0.956689\n",
       "                 bllip-xs  Center Embedding              0.617710\n",
       "                           Garden-Path Effects           0.495728\n",
       "                           Licensing                     0.120851\n",
       "                           Long-Distance Dependencies    0.445607\n",
       "                           Transformations               0.910621\n",
       "Name: correct, dtype: float64"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joined_data_circuits.groupby([\"model_name\", \"corpus\", \"circuit\"]).correct.mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Variance in accuracy vs variance in perplexity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 479.5x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "catplot_ticks = [\"correct\", \"test_ppl\"]\n",
    "catplot_data = joined_data.copy()\n",
    "catplot_data[\"correct\"] *= 100\n",
    "catplot_data = catplot_data.melt(id_vars=set(catplot_data.columns) - set(catplot_ticks))\n",
    "catplot_data[\"corpus_size\"] = catplot_data.corpus.map(corpus_to_size)\n",
    "\n",
    "g = sns.catplot(data=catplot_data,\n",
    "                x=\"variable\", y=\"value\", hue=\"model_name\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Circuit–circuit correlations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:10: FutureWarning: using a dict on a Series for aggregation\n",
      "is deprecated and will be removed in a future version\n",
      "  # Remove the CWD from sys.path while we load stuff.\n",
      "/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:11: FutureWarning: using a dict on a Series for aggregation\n",
      "is deprecated and will be removed in a future version\n",
      "  # This is added back by InteractiveShellApp.init_path()\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,0.98,'Circuit--circuit correlations')"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1440 with 25 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f, axs = plt.subplots(len(circuit_order), len(circuit_order), figsize=(20, 20))\n",
    "plt.subplots_adjust(hspace=0.5, wspace=0.5)\n",
    "\n",
    "for c1, row in zip(circuit_order, axs):\n",
    "    for c2, ax in zip(circuit_order, row):\n",
    "        if c1 <= c2:\n",
    "            ax.axis(\"off\")\n",
    "            continue\n",
    "            \n",
    "        xs = results_df[results_df.circuit == c1].groupby([\"model_name\", \"corpus\", \"seed\"]).correct.agg({c1: \"mean\"})\n",
    "        ys = results_df[results_df.circuit == c2].groupby([\"model_name\", \"corpus\", \"seed\"]).correct.agg({c2: \"mean\"})\n",
    "        df = pd.concat([xs, ys], axis=1)\n",
    "        ax.set_title(\"%s /\\n %s\" % (c1, c2))\n",
    "        sns.regplot(data=df, x=c1, y=c2, ax=ax)\n",
    "        \n",
    "plt.suptitle(\"Circuit--circuit correlations\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Stability to modification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Stability to modification')"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(15, 10))\n",
    "sns.barplot(data=results_df_mod, x=\"model_name\", y=\"correct\", hue=\"has_modifier\")\n",
    "plt.title(\"Stability to modification\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Stability to modification')"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplots(figsize=(15, 10))\n",
    "sns.barplot(data=results_df_mod, x=\"corpus\", y=\"correct\", hue=\"has_modifier\")\n",
    "plt.title(\"Stability to modification\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:1: FutureWarning: using a dict on a Series for aggregation\n",
      "is deprecated and will be removed in a future version\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Change in accuracy due to modification')"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "avg_mod_results = results_df_mod.groupby([\"model_name\", \"test_suite_base\", \"has_modifier\"]).correct.agg({\"correct\": \"mean\"}).sort_index()\n",
    "avg_mod_diffs = avg_mod_results.xs(True, level=\"has_modifier\") - avg_mod_results.xs(False, level=\"has_modifier\")\n",
    "\n",
    "plt.subplots(figsize=(15, 10))\n",
    "sns.boxplot(data=avg_mod_diffs.reset_index(), x=\"model_name\", y=\"correct\")\n",
    "plt.title(\"Change in accuracy due to modification\")"
   ]
  },
  {
   "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
}