\n",
"\n",
"\n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
"

\n",
"

"
],
"text/plain": [
" Review \\\n",
"0 The best show I've seen so far this year! \n",
"1 Really enjoyed it \n",
"2 Amazing. \n",
"3 Why all the hate? I enjoyed it. \n",
"4 Beautiful visuals, entertaining, and I believe this show has a lot of potential! \n",
"5 A beautiful rendering of Middle Earth's history \n",
"6 So far, so good... and there's still hope \n",
"7 I'm a fan \n",
"8 It works for me \n",
"9 Not the best, but enjoyed every episode. Can't wait to see much much more. \n",
"10 Beautiful, flawed, and a wonderful Fall treat \n",
"11 Good show with too many subplots \n",
"12 Starts badly, gets better \n",
"13 Good and bad things \n",
"14 Big and beautiful but can use a little help with its energy. \n",
"15 Pretty but ultimately hollow and lacking in engagement \n",
"16 Beautiful visuals and story overshadowed by unnecessary gore and violence \n",
"17 Not what you're probably expecting \n",
"18 Poor writing; Uninteresting characters, nonsensical actions. \n",
"19 Budget was spent on snacks between shots \n",
"20 If you ignore the source material, it's still boring and weird \n",
"21 Just a bad show \n",
"22 It's awful \n",
"23 I was hopeful... \n",
"24 Painfully mediocre with a few good spots \n",
"25 Beautiful to look at... but that’s about it. \n",
"26 Underwhelming and disappointing \n",
"27 Tolkien is rolling in his grave. No mystery. No inspiration. Wardrobe & acting is pretty bland. \n",
"\n",
" Label Prediction (GPT-2) Prediction (GPT-Neo) Prediction (GPT-J) \n",
"0 positive positive positive positive \n",
"1 positive positive positive positive \n",
"2 positive positive positive positive \n",
"3 positive positive negative positive \n",
"4 positive positive positive positive \n",
"5 positive positive positive positive \n",
"6 positive positive positive positive \n",
"7 positive positive positive positive \n",
"8 positive positive positive positive \n",
"9 positive positive negative positive \n",
"10 positive positive positive positive \n",
"11 neutral neutral negative negative \n",
"12 neutral neutral negative positive \n",
"13 neutral neutral negative neutral \n",
"14 neutral positive neutral negative \n",
"15 neutral negative negative negative \n",
"16 neutral positive positive negative \n",
"17 negative neutral negative negative \n",
"18 negative positive negative negative \n",
"19 negative neutral negative negative \n",
"20 negative positive negative negative \n",
"21 negative neutral negative negative \n",
"22 negative negative negative negative \n",
"23 negative positive neutral negative \n",
"24 negative neutral neutral negative \n",
"25 negative positive neutral negative \n",
"26 negative negative negative negative \n",
"27 negative positive negative negative "
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"pd.set_option('display.max_rows', None)\n",
"pd.set_option('display.max_colwidth', None)\n",
"display(pd.DataFrame(table))"
]
},
{
"cell_type": "code",
"execution_count": 22,
"id": "bed1e0e3",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"Review | Label | Prediction (GPT-2) | Prediction (GPT-Neo) | Prediction (GPT-J) | |
---|---|---|---|---|---|

0 | The best show I've seen so far this year! | positive | positive | positive | positive |

1 | Really enjoyed it | positive | positive | positive | positive |

2 | Amazing. | positive | positive | positive | positive |

3 | Why all the hate? I enjoyed it. | positive | positive | negative | positive |

4 | Beautiful visuals, entertaining, and I believe this show has a lot of potential! | positive | positive | positive | positive |

5 | A beautiful rendering of Middle Earth's history | positive | positive | positive | positive |

6 | So far, so good... and there's still hope | positive | positive | positive | positive |

7 | I'm a fan | positive | positive | positive | positive |

8 | It works for me | positive | positive | positive | positive |

9 | Not the best, but enjoyed every episode. Can't wait to see much much more. | positive | positive | negative | positive |

10 | Beautiful, flawed, and a wonderful Fall treat | positive | positive | positive | positive |

11 | Good show with too many subplots | neutral | neutral | negative | negative |

12 | Starts badly, gets better | neutral | neutral | negative | positive |

13 | Good and bad things | neutral | neutral | negative | neutral |

14 | Big and beautiful but can use a little help with its energy. | neutral | positive | neutral | negative |

15 | Pretty but ultimately hollow and lacking in engagement | neutral | negative | negative | negative |

16 | Beautiful visuals and story overshadowed by unnecessary gore and violence | neutral | positive | positive | negative |

17 | Not what you're probably expecting | negative | neutral | negative | negative |

18 | Poor writing; Uninteresting characters, nonsensical actions. | negative | positive | negative | negative |

19 | Budget was spent on snacks between shots | negative | neutral | negative | negative |

20 | If you ignore the source material, it's still boring and weird | negative | positive | negative | negative |

21 | Just a bad show | negative | neutral | negative | negative |

22 | It's awful | negative | negative | negative | negative |

23 | I was hopeful... | negative | positive | neutral | negative |

24 | Painfully mediocre with a few good spots | negative | neutral | neutral | negative |

25 | Beautiful to look at... but that’s about it. | negative | positive | neutral | negative |

26 | Underwhelming and disappointing | negative | negative | negative | negative |

27 | Tolkien is rolling in his grave. No mystery. No inspiration. Wardrobe & acting is pretty bland. | negative | positive | negative | negative |

"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"

"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAT8AAAEMCAYAAACySLGoAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAAsTAAALEwEAmpwYAAAj30lEQVR4nO3deZyVZf3/8deAuCWLRqaACoh8MtPIQEvNpVIT9yx/rmhZbqWS5ZJfFcncyAVJcym3EDXKJTfSn5o7fN01BT/uigOWiCOu4cyc7x/XPXAYD8x9zpxz3Wd5P3mcx8y573Pf12fux8zFtV9NuVwOEZFG0yPrAEREsqDMT0QakjI/EWlIyvxEpCEp8xORhqTMT0Qa0nJZB1ACjc0Rqaym7t7g03mvpP477dV/aLfTK0UtZn58Ou+VrEOoWr36D2W55QdmHUZVa13YrGe0DK0Lm7t/k/a27t+jwmoy8xORKtfWmnUEXVLmJyJll8u1Zx1Cl5T5iUj5tSvzE5FGpJKfiDQkdXiISENSyU9EGlFOvb0i0pDU4SEiDUnVXhFpSOrwEJGGpJKfiDQkdXiISENSh4eINKJcTm1+ItKIKtDmZ2ZnA3sAg4EN3f3Z5Phw4Crg88A7wBh3f7Gr+0VbydnMVjezq83s/uT9RmZ2aKz0RSSi9vb0r/RuArYEXu90/GLgQncfDlwIXJLmZjGXsf8j8CDQL3n/PHB4xPRFJJZce/pXSu7+oLvPzj9mZqsDGwPXJoeuBTY2sy90db+Y1d6B7n6xmR0C4O4Lzaz6W0VFpHhtn6b+qJn1Y3GhKF+Lu7d0cflaQLO7twG4e5uZzUmOv72sC2OW/Jbo+05+4EzW7heRCiuu2jsWeLXAa2wlQ4xZ8rvBzC4BepvZgYQq7+UR0xeRWIrr8JgIXFngeEuKa2cDA82sZ1Lq6wkMSI4vU7TMz90nmNm+hOLtaGCSu18dK30RiaiIjoykattSSjLu/h8zewrYG7g6+fqkuy+zygsRMz8zW8fdpwBTYqUpIhmpwCBnM5sEfB9YA7jLzN5x9w2AQ4GrzOxk4F1gTJr7xaz2Pmxms4ArgOvd/ZOIaYtIRLkiOjzScvcjgSMLHH8e2LTY+8Xs8FibULffHZhtZpea2Tcjpi8isVRgqEu5xWzzawNuBW41s9WAMwjj/nrGikFEItHc3iUlmd4+wAFAX+DkmOmLSCRa0moxM7sB2IIwRWWsuz8UK20RiUwlvyVcD+zr7h9HTFNEsqCSH5jZCu7+X+BGoMnMVs4/7+4fVToGEYmstfoXM43R2zs9+foB8H6BrzXjxNPPZcsd92K3/RYvRnPHPQ+w676HsOEWo3l21gsZRld9tt9ua5579n6en/kgxx7zs6zDqTp1/XzU2wvuvnHyNeawmorYbfS27LPHLpxw6tmLjg0bug4TTz+J8b+blGFk1adHjx5MOv80vjd6b958cy4zpt/OLbfeyaxZXS6z1hDq/vnUQJtfzPX8JqY5Vs1GjtiQvn16L3Fs3cFrM2SdQRlFVL02GfU1Xn75NV599Q0+/fRTpk79O7vsvH3WYVWNun8+NVDyi1ka27LAsa0ipi8RDRi4BrPfnLPo/ZvNcxkwYI0MI6oudf98KrOYaVnF6PD4IbAnMNjMpuad6guos0OkHqm3F4AXgNuATZKvHRYAd0dIXzIwp/kt1ho0YNH7QQPXZM6ctzKMqLrU/fOpgd7eGB0eTwNPm9nN7j6/0ulJdXj0sacYNmwIgwevRXPzW+y5567sP6bOejS7oe6fTy6XdQRdilHtPcrdzweON7PPnHf3YysdQ7kcM+5MHn3yGVpaFvCd3fbj8IP2p2+fVTjjvIuY3/Iehx8zji+tN5RLzzst61Az19bWxlFjT+T2266hZ48eXHnVX5g5U0OBOtT986mB3t4Y1d6Opas+jJBWRf1u/PEFj393q80jR1Ibpv3jHqb9456sw6hadf18lPmBu1+SfB1f6bREpErUQIdHzHF+R5tZ3+T7yWb2vJltFyt9EYmorS39KyMxx/kd6O7vmdk2wOrAj4HTI6YvIrHUwDi/mJlfRxa/DTDF3R+OnL6IxFIDmV/MJa0+NrPjCLsrfcvMmoDlI6YvIrGozW8JBwJrAse5+1vAULSTm0hdyrXnUr+yEnMPjxeAsWb2OTP7nLu/TNjHQ0TqjYa6LGZm6xJKeiOAnJk9Cezn7q/EikFEIsmwFzetmNXei4FLgZWAlYE/ApdETF9EYlGHxxK+4O6X572/wsyOipi+iMRSA9XemCW/dsub3Gtmw1k8/EVE6kkul/6VkZglvxOAB8zsKaAJ2AjYP2L6IhJLDZT8Yvb2/sPMNgA2TQ7NcPd5sdIXkYgyHMKSVsySH0D1PxER6T719i5mZt8HngeOAI4EZprZbrHSF5F4cu3tqV9ZiVnyOw3YLBnsjJmtB9wM3BQxBhGJQdXeJXzSkfEBuPuLZvZxxPRFJJYKze01s52AUwmdpk3AeHe/oZR7xcz8/m5m/wNcRgj6R8BNZrYS0OTu2slNpF5UoOSXLIYyGfiWuz9rZhsBD5nZTe5edG4bM/M7Ofl6aqfjpxA6QnpGjEVEKqk1fYeHmfUD+hU41eLuLZ2OtRO2vSW5Zm4pGR/EHeqitftEGkVx1d6xwLgCx8cTCkcAuHvOzPYk1CI/BHoDo0sNURmSiJRfey79CyYCQwq8Jubf0syWA34N7Oru6wA7A1PNbJVSQow9zk9EGkAxQ1iSqm1Lio+OAAa4+0PJdQ8lJcD1gUeLjVElPxEpv+JKfmm9CQzqWCPAzNYHvgi8XEqIKvmJSPlVoLfX3d8ys8OAv5lZR9Hyx+4+v5T7KfMTkfKr0PQ2d59Cmba/UOYnImWX5d4caSnzE5HyU+YnIg1J6/mJSENSyU9EGpIyPxFpRLk2VXtFpBGp5FcZvfoPzTqEqta6sDnrEKqenlFlaahLhazbf+OsQ6haL897Qs+nCy/Pe4Lfr7Vf1mFUrSNmX939myjzE5GGVP1Nfsr8RKT8cq3Vn/sp8xOR8qv+vC9d5mdmXwA+dvcPzKwnMIbw400udQlpEalf9dThcStwKPAkYQvKnYFPCYsL/qIikYlI7aqBIlHazG848FTy/X7AZsAHwHMo8xORTmqh5Jd2Jec2YHkz2xB4z93fICw7XdLa+SJS59qLeGUkbclvGjAV+DxwXXLsy4BGiorIZ+Ras46ga2kzv58ABxDa+SYnx/qTt62ciEiH4nauzEaqzM/d/wtc2unYvZUISETqQC1nfmY2Geiy1dLdx5Q1IhGpebVe8nspWhQiUldqOvNz9/ExAxGR+pFra8o6hC6lnt5mZtsCewGru/vOZjYS6OPu91QsOhGpSbVQ8ks1zs/MjgAuAl4EtkwOfwz8tkJxiUgNy7U3pX5lJe0g57HAd939TBb34zwPWCWCEpHalmtP/8pK2mpvb2B28n1HD3AvYGHZIxKRmpfLVX+bX9qS3/3A8Z2OHQn8s7zhiEg9qKeS3xHALWb2U6C3mTnwPrBTVxea2YRlnXf3Y1PGICI1or1eenvdfa6ZjQJGAesQqsCPpFzL78NuxCciNSjLjoy0ilnJuQehnQ+gJ5Dqp9N4QZHGUzeZn5ltBNwErEBYyWUQ8ImZ7e7uT6dNzMy2IyyAumLHMXf/TRHxikgNyFVoOT8zWxE4D/gu8Akw3d0PLuVeaTs8LgcuBAa5+ybAQOCC5HgqZnYmcBxwdHL94YRFUkWkzlRwnN8EQqY33N03BE4qNca0md9wYKK75wCSr+cD6xWR1o7A9sC/3f0Q4OvAakVcLyI1IpdrSv1Ky8xWIewfdFJeXvTvUmNM2+Z3O7ALcGPesZ2B24pI6xN3bzWznJn1cvdmMxtUxPUiUiPaiujtNbN+QL8Cp1rcvSXv/brAO8A4M9uGsJXGie7+YCkxpl3SqidwnZk9TujpXYtQcvt7EWm9b2YrAw8DV5nZXMIUORGpM0UOch4LjCtwfDxLLpjcExgKPOnux5jZpoQheMPcfUGxMRazpNWzed/PBO4oMq29gVbgV4R2v37AD4u8h4jUgKLa8pqYCFxZ4ExLp/dvEPKQawHc/X/NbB6hWe6xYmOMsqRVstfvb939p4QpcVoQQaSOFdPb6y94C5/N6D77Ofd5ZvZPYFvgTjMbDqxOiWuPFrOk1fKEhQz6kzfGL82SVu7elgyXEZEGUMFxfocCl5vZOYQ9hfbv1C6YWtpxflsAfyWM8+sDLGDxYgdDU6Z1j5ldAPyZ0FAJgLvPLCZgEal+be1pB5IUx91fAbYux73SRngeMMHdVwPeT76eCvyhiLT2Igx3+Quhl/g24NYirq8qZ54/jkdm3cW0B6ZmHUpV0vNJp6lHE3tN+y07XfHLrEMpq1wu/Ssraau9wwnj+vKdCbwKnJ3mBu4+pIi4qt71193C5Mv+wtkXaoJKIXo+6Xz1oO8x/6U5LL/KSlmHUlbtdbSk1XuE6i7AXDP7MrAqsErahMzsM0WAQsdqxaPTn6Dl3feyDqNq6fl07XNrrMbgb49g5rX3Zh1K2VVikHO5pc38bgBGJ99fTljH73Hgb0WkNazAsfWLuF6krmx5yn48dPq15NozrPtVSN1Ue919bN73Z5vZ/xJKfV2O9UvWADwYGG5mj+Sd6gt4UdGK1InB3xnBR+8s4O1/vcbAb9RfGaAWqr3FLGm1iLs/UMTH7yRsfHQBcEze8QXAM6WkL1Lr1hw5nKHbbszgbb5KzxV6sXzvldj2/MP4/0ddlHVoZVGp3t5yWtb0tgdYPL1tqdx9yy7Ovw68Dnyl6OhE6tT0s6Yy/azQ5D3wG+vztUNG103GBykyjiqwrJLfn8qZkJk9SoFnkiyRVXMmXno6m27+dVZdrR8PPjON88+6mL9OKWaqc33T82lsNV3tdferypzWr/K+X5Ew13dOmdOIZuzBJ2QdQlXT80mvecYsmmfMyjqMsqqF3dtKavMrhbvfl//ezO4ESlqKRkSqW4absqUWLfMroA+wRobpi0iF5NJt8ZOpaJlfpza/HoQ5wefESl9E4mlVtXcJ+W1+rcAr7j43YvoiEkndlPzMbAXgZEInxefdvW+yE9twd78gzT3c/T4z6wMMc/cnSo5YRKpeLbT5FbOqy1eAfVlcdX0OOCxtQmY2OrnmhuT9SDO7JX2oIlIrcjSlfmUlbea3O7CPu08nydTdvZmwBWVa44FRwLvJ9Y8RNiQRkTrTXsQrK2kzv4V0qiKb2RcIOyml5u5vdTr032KuF5Ha0EZT6ldW0mZ+fyXsuDYEwMzWJMzVva6ItN43sy+SVJvNbGtSrNsvIrWnvSn9KytpM78TCAuX/ouw69qLhNkZxWxydDwwDRhiZvcCU1iyB1hE6kQ7TalfWUm7pNVC4BfAL5Lq7ryOHdPTcvdHko2GNyNsgPRwqRuPiEh1q/WFDRYxs86bFPU2M2DRhiKpuPt7ZnZ3R7pmtrK7f5T2ehGpDbUw1CXtIOeXCJl5fhm1I3PvmeYGZvZ9YBKwZnKoKblHqutFpHa0N9XJIGd3X6Jt0MzWAMYBxSxqOgHYE5jh7rXwH4OIlKgt6wBSKHUl57fMbCzwAnBNysvmu/vDpaQnIrUly17ctLozt9eAlYv4/I1mdhhh395POg6qzU+k/mTZi5tW2g6PzkvarwxsABSzKetpydcLWdx+qDY/kTpUN729fHZJ+w+Bp939xbQJdW43FJH6VRfVXjPrCXwbONjdNR1NRLpUCz2aXZbG3L0N2I7a+HlEpAq0NaV/ZaWYJa3Gm1mvSgYjIvWhFlZ1WWa118z2dvdrgSMI+20cbWZvk9ee6e5rVzZEEak1tVBN7KrN7xLgWmC/CLGISJ2o9BYeZjYOOAXY0N2fLeUeXWV+TfDZbSdFRJalkiU/M9sY+Abwenfu01Xm1zNZiWWp+bi739OdAESk/hQzvc3M+hGWyuuspfPKT8l+QhcS9hO6t7Togq4yvxWAy1h65pcjbEEpIrJIkeP8xhLWCuhsPKFqm+83wNXu/lrHylKl6irz+9DdlbmJSFGKrPZOBK4scLwl/42ZfRMYSVgYudti7tsrIg2imMwvqdq2pPjoVsD6wKtJqW8QcIeZ/cjd7yw2xlQdHiIixajE3F53PxM4s+O9mb0G7FSR3l53713KTUWksdXF3F4RkWLFWMzU3Qd353plfiJSdu01sKhVTWZ+L897IusQqpqeT9eOmH111iHUtXqY3laVllt+YNYhVK3Whc16Pl1oXdjMp/NSbzrYcHr17/7otuov99Vo5ici1U0lPxFpSOrtFZGG1FYDFV9lfiJSdqr2ikhD0lAXEWlI1Z/1KfMTkQpQtVdEGpI6PESkIankJyINKaeSn4g0IpX8RKQhaaiLiDSk6s/6ImZ+ZrY6cC6wtrtvaWYbAZu5+8WxYhCROFprIPvrETGtPwIPsnh/zueBwyOmLyKR5Ir4l5WYmd/ApJTXBuDuC6mNdlERKVJ7Ea+sxGzza81/k+zSXgML34hIsWphqEvMkt8NZnYJ0NvMDgTuBC6PmL6IRFILJb9omZ+7TwDuBx4HRgOT3P38WOmLSDxtuVzqV1Zi9vau4+5TgCmx0hSRbGic35IeNrNZwBXA9e7+ScS0RSQitfktaW1gIrA7MNvMLjWzb0ZMX0QiqYU2v2glP3dvA24FbjWz1YAzCOP+esaKQUTiULW3kyTT2wc4AOgLnBwzfRGJoxaqvTE7PG4AtgBuAsa6+0Ox0haRuLLsxU0rZsnvemBfd/84YpoikgFVewEzW8Hd/wvcCDSZ2cr55939o0rHICJx1cK81Rglv+nAxsAHhJVumjp9VYeHSJ2pRJufmX0emAysCywEXgQOcfe3S7lfxTM/d984+RpzWI2IZKhC1d4cMMHd7wUws98BZwIHlXKzaBmSmU1Mc6xWbL/d1jz37P08P/NBjj3mZ1mHU5X0jD7rxNPPZcsd92K3/Q5ddOyOex5g130PYcMtRvPsrBcyjK58crlc6lda7j6/I+NLzADWKTXGmB0eWxY4tlXE9MumR48eTDr/NL43em/efHMuM6bfzi233smsWS9mHVrV0DMqbLfR27LPHrtwwqlnLzo2bOg6TDz9JMb/blKGkZVXMVtXJis89StwqsXdW5ZyTQ/gMODm4qMLYnR4/BDYExhsZlPzTvUFarKzY5NRX+Pll1/j1VffAGDq1L+zy87bN/wfdj49o8JGjtiQ5rn/XuLYuoPXziiayimy2jsWGFfg+HjglKVc83tCP8IFxSSUL0bJ7wXgNmCT5GuHBcDdEdIvuwED12D2m3MWvX+zeS6bjPpahhFVHz2jxlZMdZYw7fXKAsdbCn3YzM4G1gN2dveSO5ZjdHg8DTxtZje7+/xKpyci2Sum5JdUbVvSfNbMTge+DuyYDKErWcw2vwVmdjAwAlix46C7/zhiDGUxp/kt1ho0YNH7QQPXZM6ctzKMqProGTW2Cg112QD4NaE2+bCZAbzq7ruXcr+Ymd8lSXrbABcR5vjeHzH9snn0sacYNmwIgwevRXPzW+y5567sP0a9mfn0jBpbJaa3uftzlHHri5iZ3ybuvqGZPePuZ5jZH4C/R0y/bNra2jhq7Incfts19OzRgyuv+gszZ9bHEIVy0TMq7JhxZ/Lok8/Q0rKA7+y2H4cftD99+6zCGeddxPyW9zj8mHF8ab2hXHreaVmH2i21ML2tqciGyZKZ2SPuvomZPQls7u4fmdlMd/9ykbfKLbf8wEqEWBdaFzaj57NsrQub+XTeK1mHUbV69R/a7dLVNwdukzpjmd78z0w2MotZ8ptvZqsC/wCmmdk8oDli+iISSaxCVXfEnHK2o7u/C/wP8Cfgn8AeEdMXkUjayaV+ZSX2Ss4k43Imx0pXROLTYqZ5zOxt+MwTeY+w6sux7q5xECJ1oi1X/YtaxWzzu5Awf+9yQnf1GKCVMMXtUmCXiLGISAXVQptfzMxvB3ffNO/9L83sUXcfZWbPRYxDRCqsFoa6xOzwWDXZwAhYtDBhn+TtwohxiEiF5Yr4l5WYJb9JhDm+tyfvdwAmmNkqgDYzEqkj7ar2LubuF5jZ/Sxew+9Cd38m+f7nseIQkcpTb+9nvQYs5+5PRE5XRCKqhd7emMvYjwaeA25I3o80s1tipS8i8bTncqlfWYnZ4TEeGAW8C+DujxF2YRKROlMLHR5Rd1QrMJC5W4sRikh1qoWSX8w2v/fN7IskszzMbGtSrt4qIrVFHR5LOh6YBgwxs3sJa/BrVodIHWrLtWUdQpdiDnV5xMy2ATYjTG97eGnb0olIbauF6W2x2/zeI+zYdi+w0MxWjpm+iMShJa3ymNn3CbM81kwONRHa/3rGikFE4qiFkl/MNr8JhM3LZ3Rnr00RqX6a3rak+e7+cMT0RCQj6u1d0o1mdhjwF+CTjoPu/lHEGEQkglqY3hYz8+vYi+9CQluf2vxE6pTa/PK4e9SeZRHJjtr8RKQhqeQnIg2pFpaxV+YnImWnkp+INCT19opIQ1KHh4g0JFV7RaQhVWqGh5kNB64CPg+8A4xx9xdLuZfG3olI2eVyudSvIl1M2PlxOGHCxCWlxqiSn4iUXTFtfmbWD+hX4FRL/pqfZrY6sDGwbXLoWuACM/uCu79dbIy1mPk1tS5szjqGqqbn07Ve/YdmHUJda13Y3JT2s2Z2CjCuwKnxwCl579cCmt29DcDd28xsTnK8ITI/EakvE4ErCxxvqWSiyvxEJFNJ1bYlxUdnAwPNrGdS6usJDEiOF00dHiJSE9z9P8BTwN7Job2BJ0tp7wNoqoXxOCIiAGb2JcJQl1WBdwlDXbyUeynzE5GGpGqviDQkZX4i0pCU+YlIQ1LmJyINSZlficzsUDP7RfL9CDPbs9P5p8xspWyiqz5mNtjMDu7G9aeY2dnljKla6fcpDg1yLpG7X5z3dgSwEzA17/yIyCFVu8HAwcClhU6a2XLu3ho1ouo1Av0+VVxDDXUxsxzwG2BXYCXgBHe/Pjn3PeAMwlaabwOHuPtLZmaEqTcrJ+eudPezk/mIqyTXPAn0AV4D7nf3I5O0egO7AXu4++5JOssBbwCbu/urZnYcsAfhP6Jm4Kfu/laFH0WXkvj/B9idsHzQMXnPalPgTMLPDHCyu99mZlsDZ7v7yORzi96b2XPAEOAF4CV3/4GZvQZcB3wb+FeS3rXJfVcEbnP3Y5N7nQKs4u6/qvCPXlApzyM593PgKMIMhtuBn7l7/+T34LbkXisBjwCHEH5n6u73qRo1YrW3LflfdBfgUjNbPVktYjKwr7tvBFwDTEk+fzhws7t/1d2/AlyWfzN3fwc4GbjL3Ue4+5Gd0rsB+JaZ9U/e7wA8n/yi7gesC3zD3Tcm/HGcU+4fuBsWuPsoYH9gEixageNiYB93/zqhhHJJcnxZfgbMTJ7RD/KO93H3Tdz9IEIGsXNy3xHAyOQ/pWpR1PMws42AXwObJdf1y7tXW3LNSOArhP9Yf1znv09VpRGrvZcBuLub2RPANwibpz/t7jOTz1wB/MHMegP3AxPMbGXgn8krNXf/yMxuAvYh/MEcyOJJ3LsAI4EnQgGT5YD3Sv3BKuC65OsMYICZrQhsRijBTUtihvD8hpWYxp/zvu8J/M7MNiNsar8GIRP8R4n3Lrdin8dmwO15068uB/ZNvu8B/MrMdiD83KsCH3UVQI3/PlWVRsz8iuLu15vZdGA74Hjgx8B+Rd7mSuB8M5sCbEUoOUD4A/+tu19epnDL7RNYtHQQhN+XJuAZd9+y84fNbAuWrE2smCKND/K+P5qQCWzq7p+Y2aUp7xFLsc9js2Xcax9gC+Bb7v6+mZ0ADE8Zx5XU5u9TVWnEau+PAMxsPeBrhP/FZwBfTeYNAhxAmDD9vpkNA95y9ysJ64ttUuCeC4C+S0vQ3R8ktOGcAdzk7h3/w98MHG5mqyYxrWBmX+3mz1dpDwPrmdk2HQfMbJSZNQGvAEPNbNXk/d551y3zGSX6AXOTjG8goW222i3redwH7JBXRT0g77p+wLzkd6wvITPs0Ei/T5lpxMxvOTN7EriV0Knxn6Rasj9wjZk9QyjZdZTu9gT+lVzze0LjdWd3A58zs6fNbNJS0r0K+Cl565a5+2RC2+J9SbqPA5t39wesJHd/l1C9Gpf8vLMIC042ufscQhvT44RMYW7epc8AbmbPmtnflnL7ScDmZvYsoXni7gr9GGXTxfN4GpgATDezx4FWFldD/wz0NrPngVuAB/Ju2zC/T1lqxN7e3u7+QZcfFikDM+vt7u8n358CDHP3YptNpALU5idSWWea2ebA8oRmgZIHekt5NVTJT0SkQyO2+YmIKPMTkcakzE9EGpIyvyqSrHySS+ZrdvXZA83swRhxFUh7iTjNbJqZHdDVdQXus7aZfWBhF65ozGz7ZJZE5szs+mSWh0Sm3t4SJZPyBwAD3H1e3vEnCVOyhrj7a5kEF5m7p/rjTZ7ZT9z9ruS6NwiLQ8R2GvDzvLiaCHOPDyZMS1sAPA9c7O7XJZ+5lzAVspUw0+P+5JrLgW8lt1qBMLVtYfL+anc/ND/hZLGHq919UHLoLOAiYFo5f0Dpmkp+3fMqebMYzGxDwuovNSVNSbNemNkooK+7z8g7PAkYC/ySsMrKQOBEoPOiCj9391UI09D6Aee5+w7uvkpyfAowoeN954yvEHd/BOhjZiO7+aNJkRrml75CJgNjCDM/IExf+jPw244PJFOXfk9YfeMj4I/A6e7enlT3ziJMTl9ApxU4kmvPBUYD7YQFF8a5e9uygjKzwYSM+RCS2QbAOe5+dnL+FMJKIp8QZiccbWZ/XVpaKeK8l1Ca+VPy/qeEebqDCBtK7wf8AlgbuMXM2ghLi01N4uzl7q1mNoCwQsoWwHzgLHf/Y17MX05i3p2wjNMB7v5Ycv444EjCtK85wOHuXmiGyA6EaWcdsQ8nrNyzace9Eg8mr89w9/lmdj1wWKHzJbgX2BF4rIvPSRmp5Nc9Mwj/a6+fZBB7AVd3+szvCfM0hxImoY8hmV9MmJ60E2GO8UjgB52uvZJQzRqWfGY74CdFxLcNsF5y3XFm9t28c7sCfyOUYKZ0kVZXcS5iZj8kZLhjCBnRLsA77r4/IcPaOSkVTShw+XXAm4TmhB8Ap5vZt/PO75J8ph9hHusFSZpGqMaOcvfewPaEtfAK2RDI3+f128DsThnfMiVzdfcgrLtXDrMAzcGNTCW/7uso/d1H+CVu7jiRlyGOSKY4vW9m5xDmEV9GmDc80d1nJ58/A9g6+f6LhFJYP3f/GPjQzM4jtEtdkjK28e7+IWFu8hWEKvpdybnp7n5TklafLtJaapwF/IRQ9Xs0ef9SmkDNbC3CPNQd3f0T4Ckz+xPh2d6TfOxBd789+fxkQlUVwtp4KwBfNrO3u2hr7Qe8n/e+P7DEYp9m9iahLXJFwNz99eTUJAtL6X9IKK0dneZnS+F9llzrTyJQ5td9kwmN30NYcm06CH9YvYDX8469TmhTglDCmd3pXId1kmvn5q0T16PT57vS+d4bLuVcV2ktK87O1gJeLiLGDgOA+R3zYPPSyW8Ly8+kPgJWtLD8/UtmNpZQ4tzAzO4Ajk4WWujsXcKKyB3eAdbM/4C7D0raQT8lNBl0OLKjap+GhdWr10ne7uDuDyzlo70JC7lKRMr8usndXzezVwklp4M6nZ5H+ANaB+hYKHVtFpcO5xIyC/LOdZgN/Bfo76XvbbEWodey4975mUH+vMau0lpWnJ3NJqwmXMiy5lLOAVbLXwiAJZ/VMrn7NYRVefoQSqtnsXidu3zPsOS6efcAF5jZyGKqvilj2iDlR9cHni5n2tI1ZX7lcRCwqrt/mN9zmnQWTAVOM7MxwGqEqlLHLmRTgSPN7FZCVer4vGvnmtmdwDlmdhJh0c8hwCB3X9Rg34WTks6HIYR2xoKriaRIa6lxFvAn4NxkDOIThIzw06Tq+G9C22ehGGab2cPAGWb2K0IGdRCLVz5eqqTNbyDwEKFD5GPC6siF3M7iFZk7VvS+BLjOzA4jdHIsJKzCHMtWFL9ArnSTOjzKwN1fXkap4QhChvEK4Q/rGsLYMAg9v3cQ/td/grA/Q74xhNVAZhKqa3+jUxWtC/cR2tzuJmwkdOcyPrustLqKcxF3/ythHN01hLasmwiZPoTFN080s5Ykg+tsb8Iub3OAGwm9zXcV+FxnKxA2EJpHqBqvTtg7o1B8TwDvWdh0qMPPCMNdziX0Mr8JnAr8P0InTbktKgEnQ28+SIa8SERa1aUO5Q116dWNKnPdMrPtCENhdssg7V2A33iyFWUyZOayjo4ciUfVXmk4SQl4WaXgikiaRPYgbzyfu+8ROw4JlPmJRJAMWJ9NWFp+TMbhCKr2ikiDUoeHiDQkZX4i0pCU+YlIQ1LmJyINSZmfiDQkZX4i0pD+Dwj5YYBsmY4cAAAAAElFTkSuQmCC\n",
"text/plain": [
"

"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"for model_name, model_predictions in predictions.items():\n",
" mat = confusion_matrix(test_labels, model_predictions, labels=labels)\n",
"\n",
" sns.heatmap(mat, square=True, annot=True, xticklabels=labels, yticklabels=labels, linewidths=1)\n",
" plt.xlabel(f\"Model predictions ({model_name})\")\n",
" plt.ylabel(\"True labels\")\n",
" plt.show()"
]
},
{
"cell_type": "markdown",
"id": "bc526679",
"metadata": {},
"source": [
"# Chain of thought"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "a7361925",
"metadata": {},
"outputs": [],
"source": [
"prompt = '''Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?'''"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "f9c0a171",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: The answer is 6.\n",
"Q\n",
"-------------------------\n",
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: The answer is 6.\n",
"Q\n",
"-------------------------\n",
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: The answer is 17.\n",
"Q\n",
"-------------------------\n"
]
}
],
"source": [
"for model_name in models:\n",
" print(generate(prompt, model_name=model_name, max_new_tokens=10))\n",
" print('-------------------------')"
]
},
{
"cell_type": "code",
"execution_count": 25,
"id": "a3985e6c",
"metadata": {},
"outputs": [],
"source": [
"prompt = '''Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: Roger started with 5 balls. 2 cans of 3 tennis balls each is 6 tennis balls. 5 + 6 = 11. The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?'''"
]
},
{
"cell_type": "code",
"execution_count": 26,
"id": "031f7a56",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: Roger started with 5 balls. 2 cans of 3 tennis balls each is 6 tennis balls. 5 + 6 = 11. The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: 23 = 6 + 6 = 11.\n",
"Q: Roger has a tennis ball.\n",
"-------------------------\n",
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: Roger started with 5 balls. 2 cans of 3 tennis balls each is 6 tennis balls. 5 + 6 = 11. The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: The answer is 23.\n",
"Q: The cafeteria had 23 apples. If they used\n",
"-------------------------\n",
"Q: Roger has 5 tennis balls. He buys 2 more cans of tennis balls. Each can has 3 tennis balls. How many\n",
"tennis balls does he have now?\n",
"A: Roger started with 5 balls. 2 cans of 3 tennis balls each is 6 tennis balls. 5 + 6 = 11. The answer is 11.\n",
"Q: The cafeteria had 23 apples. If they used 20 to make lunch and bought 6 more, how many apples do they have?\n",
"A: The cafeteria had 23 apples. 20 + 6 = 26. The answer is 26.\n",
"-------------------------\n"
]
}
],
"source": [
"for model_name in models:\n",
" print(generate(prompt, model_name=model_name, max_new_tokens=20))\n",
" print('-------------------------')"
]
},
{
"cell_type": "code",
"execution_count": 27,
"id": "6d58b8ff",
"metadata": {},
"outputs": [],
"source": [
"def construct_prompt_cot(review):\n",
" return f'''Q: What is the sentiment of this movie review? \"Beautifully, thoughtfully made\"\n",
"A: The reviewer is very happy about how the movie was made. The sentiment is positive.\n",
"Q: What is the sentiment of this movie review? \"Boring, even for generic fantasy\"\n",
"A: The reviewer is bored by the movie. The sentiment is negative.\n",
"Q: What is the sentiment of this movie review? \"It's fine - as in OK - as in mediocre\"\n",
"A: The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral.\n",
"Q: What is the sentiment of this movie review? \"{review}\"'''"
]
},
{
"cell_type": "code",
"execution_count": 28,
"id": "5326d4b6",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 28/28 [02:05<00:00, 4.49s/it]"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Accuracy:\n",
"GPT-2\t57.1%\n",
"GPT-Neo\t57.1%\n",
"GPT-J\t75.0%\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"\n"
]
}
],
"source": [
"from sklearn.metrics import confusion_matrix, accuracy_score\n",
"from IPython.display import display\n",
"\n",
"predictions = {model_name: [] for model_name in models}\n",
"full_predictions = {model_name: [] for model_name in models}\n",
"table = []\n",
"\n",
"for review, label in tqdm(zip(test_reviews, test_labels), total=len(test_reviews)):\n",
" prompt = construct_prompt_cot(review)\n",
" \n",
" row = {\n",
" 'Review': review,\n",
" 'Label': label,\n",
" }\n",
"\n",
" for model_name in models:\n",
" res = generate(prompt, model_name=model_name, max_new_tokens=30)\n",
" full_prediction = res.split('\\n')[7].split('A: ')[1]\n",
" prediction = full_prediction.split()[-1].strip('.')\n",
" \n",
" if prediction not in labels:\n",
" prediction = 'neutral'\n",
" \n",
" predictions[model_name].append(prediction)\n",
" full_predictions[model_name].append(full_prediction)\n",
" row[f'Prediction ({model_name})'] = prediction\n",
" row[f'Full prediction ({model_name})'] = full_prediction\n",
" \n",
" table.append(row)\n",
"\n",
"print('Accuracy:')\n",
"for model_name, model_predictions in predictions.items():\n",
" print(f'{model_name}\\t{accuracy_score(test_labels, model_predictions) * 100:.1f}%')"
]
},
{
"cell_type": "code",
"execution_count": 29,
"id": "3da28c06",
"metadata": {
"scrolled": false
},
"outputs": [
{
"data": {
"text/html": [
"

\n",
"\n",
"\n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
" \n",
"

\n",
"

"
],
"text/plain": [
" Review \\\n",
"0 The best show I've seen so far this year! \n",
"1 Really enjoyed it \n",
"2 Amazing. \n",
"3 Why all the hate? I enjoyed it. \n",
"4 Beautiful visuals, entertaining, and I believe this show has a lot of potential! \n",
"5 A beautiful rendering of Middle Earth's history \n",
"6 So far, so good... and there's still hope \n",
"7 I'm a fan \n",
"8 It works for me \n",
"9 Not the best, but enjoyed every episode. Can't wait to see much much more. \n",
"10 Beautiful, flawed, and a wonderful Fall treat \n",
"11 Good show with too many subplots \n",
"12 Starts badly, gets better \n",
"13 Good and bad things \n",
"14 Big and beautiful but can use a little help with its energy. \n",
"15 Pretty but ultimately hollow and lacking in engagement \n",
"16 Beautiful visuals and story overshadowed by unnecessary gore and violence \n",
"17 Not what you're probably expecting \n",
"18 Poor writing; Uninteresting characters, nonsensical actions. \n",
"19 Budget was spent on snacks between shots \n",
"20 If you ignore the source material, it's still boring and weird \n",
"21 Just a bad show \n",
"22 It's awful \n",
"23 I was hopeful... \n",
"24 Painfully mediocre with a few good spots \n",
"25 Beautiful to look at... but that’s about it. \n",
"26 Underwhelming and disappointing \n",
"27 Tolkien is rolling in his grave. No mystery. No inspiration. Wardrobe & acting is pretty bland. \n",
"\n",
" Label Prediction (GPT-2) \\\n",
"0 positive positive \n",
"1 positive positive \n",
"2 positive positive \n",
"3 positive neutral \n",
"4 positive positive \n",
"5 positive positive \n",
"6 positive positive \n",
"7 positive positive \n",
"8 positive positive \n",
"9 positive positive \n",
"10 positive positive \n",
"11 neutral negative \n",
"12 neutral negative \n",
"13 neutral positive \n",
"14 neutral positive \n",
"15 neutral negative \n",
"16 neutral negative \n",
"17 negative positive \n",
"18 negative neutral \n",
"19 negative negative \n",
"20 negative positive \n",
"21 negative neutral \n",
"22 negative negative \n",
"23 negative positive \n",
"24 negative negative \n",
"25 negative negative \n",
"26 negative negative \n",
"27 negative negative \n",
"\n",
" Full prediction (GPT-2) \\\n",
"0 The reviewer is very happy about the movie. The sentiment is positive. \n",
"1 The reviewer is very happy about how the movie was made. The sentiment is positive. \n",
"2 The reviewer is very happy about the movie. The sentiment is positive. \n",
"3 The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. \n",
"4 The reviewer is very happy about the movie. The sentiment is positive. \n",
"5 The reviewer is happy about the movie. The sentiment is positive. \n",
"6 The reviewer is very happy about the movie. The sentiment is positive. \n",
"7 The reviewer is a fan of the movie. The sentiment is positive. \n",
"8 The reviewer is happy with the movie. The sentiment is positive. \n",
"9 The reviewer is very happy about the movie. The sentiment is positive. \n",
"10 The reviewer is very happy about the movie. The sentiment is positive. \n",
"11 The reviewer is not happy with the movie. The sentiment is negative. \n",
"12 The reviewer is not happy with the movie. The sentiment is negative. \n",
"13 The reviewer is happy about the movie. The sentiment is positive. \n",
"14 The reviewer is happy about the movie. The sentiment is positive. \n",
"15 The reviewer is not particularly happy with the movie. The sentiment is negative. \n",
"16 The reviewer is not happy with the movie. The sentiment is negative. \n",
"17 The reviewer is not sure what the movie is about. The sentiment is positive. \n",
"18 The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. \n",
"19 The reviewer is not happy about the movie. The sentiment is negative. \n",
"20 The reviewer is not sure about the movie. The sentiment is positive. \n",
"21 The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. \n",
"22 The reviewer is very unhappy about the movie. The sentiment is negative. \n",
"23 The reviewer is hopeful about the movie. The sentiment is positive. \n",
"24 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"25 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"26 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"27 The reviewer is not impressed by the movie. The sentiment is negative. \n",
"\n",
" Prediction (GPT-Neo) \\\n",
"0 positive \n",
"1 positive \n",
"2 positive \n",
"3 neutral \n",
"4 positive \n",
"5 positive \n",
"6 positive \n",
"7 positive \n",
"8 positive \n",
"9 negative \n",
"10 positive \n",
"11 positive \n",
"12 positive \n",
"13 positive \n",
"14 positive \n",
"15 negative \n",
"16 positive \n",
"17 neutral \n",
"18 negative \n",
"19 positive \n",
"20 negative \n",
"21 negative \n",
"22 negative \n",
"23 positive \n",
"24 negative \n",
"25 positive \n",
"26 negative \n",
"27 negative \n",
"\n",
" Full prediction (GPT-Neo) \\\n",
"0 The reviewer is very happy about the show. The sentiment is positive. \n",
"1 The reviewer enjoyed the movie. The sentiment is positive. \n",
"2 The reviewer is very happy about how the movie was made. The sentiment is positive. \n",
"3 The reviewer is not sure why all the hate. The sentiment is neutral. \n",
"4 The reviewer is very happy about the visuals, the entertainment, and the potential of the show. The sentiment is positive. \n",
"5 The reviewer is very happy about how the movie was made. The sentiment is positive. \n",
"6 The reviewer is very hopeful about the future of the movie. The sentiment is positive. \n",
"7 The reviewer is a fan of the movie. The sentiment is positive. \n",
"8 The reviewer likes the movie. The sentiment is positive. \n",
"9 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"10 The reviewer is very happy about the movie. The sentiment is positive. \n",
"11 The reviewer is happy with the movie. The sentiment is positive. \n",
"12 The reviewer is not happy with the movie. The sentiment is positive. \n",
"13 The reviewer is happy about the good things in the movie. The sentiment is positive. \n",
"14 The reviewer is very happy about the movie. The sentiment is positive. \n",
"15 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"16 The reviewer is very happy about the visuals and story. The sentiment is positive. \n",
"17 The reviewer is not what you are probably expecting. The sentiment is neutral. \n",
"18 The reviewer is disappointed by the writing. The sentiment is negative. \n",
"19 The reviewer is very happy about the budget. The sentiment is positive. \n",
"20 The reviewer is bored by the source material. The sentiment is negative. \n",
"21 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"22 The reviewer is very disappointed by the movie. The sentiment is negative. \n",
"23 The reviewer was hopeful about the movie. The sentiment is positive. \n",
"24 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"25 The reviewer is very happy about the movie. The sentiment is positive. \n",
"26 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"27 The reviewer is not particularly happy about the movie. The sentiment is negative. \n",
"\n",
" Prediction (GPT-J) \\\n",
"0 positive \n",
"1 positive \n",
"2 positive \n",
"3 positive \n",
"4 positive \n",
"5 positive \n",
"6 positive \n",
"7 positive \n",
"8 positive \n",
"9 positive \n",
"10 positive \n",
"11 positive \n",
"12 neutral \n",
"13 positive \n",
"14 positive \n",
"15 negative \n",
"16 positive \n",
"17 negative \n",
"18 negative \n",
"19 positive \n",
"20 negative \n",
"21 negative \n",
"22 negative \n",
"23 positive \n",
"24 negative \n",
"25 negative \n",
"26 negative \n",
"27 negative \n",
"\n",
" Full prediction (GPT-J) \n",
"0 The reviewer is very happy about the movie. The sentiment is positive. \n",
"1 The reviewer is very happy about the movie. The sentiment is positive. \n",
"2 The reviewer is very happy about the movie. The sentiment is positive. \n",
"3 The reviewer is happy about the movie. The sentiment is positive. \n",
"4 The reviewer is very happy about the movie. The sentiment is positive. \n",
"5 The reviewer is very happy about how the movie was made. The sentiment is positive. \n",
"6 The reviewer is optimistic about the movie. The sentiment is positive. \n",
"7 The reviewer is a fan of the movie. The sentiment is positive. \n",
"8 The reviewer is happy with the movie. The sentiment is positive. \n",
"9 The reviewer is happy about the movie. The sentiment is positive. \n",
"10 The reviewer is very happy about the movie. The sentiment is positive. \n",
"11 The reviewer is happy with the movie. The sentiment is positive. \n",
"12 The reviewer is not sure about the movie. The sentiment is neutral. \n",
"13 The reviewer is happy about the movie. The sentiment is positive. \n",
"14 The reviewer is happy with the movie. The sentiment is positive. \n",
"15 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"16 The reviewer is very happy about the movie. The sentiment is positive. \n",
"17 The reviewer is not happy with the movie. The sentiment is negative. \n",
"18 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"19 The reviewer is happy about the movie. The sentiment is positive. \n",
"20 The reviewer is not impressed by the movie. The sentiment is negative. \n",
"21 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"22 The reviewer is very unhappy about the movie. The sentiment is negative. \n",
"23 The reviewer is hopeful about the movie. The sentiment is positive. \n",
"24 The reviewer is not happy with the movie. The sentiment is negative. \n",
"25 The reviewer is not impressed by the movie. The sentiment is negative. \n",
"26 The reviewer is disappointed by the movie. The sentiment is negative. \n",
"27 The reviewer is not happy with the movie. The sentiment is negative. "
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"pd.set_option('display.max_rows', None)\n",
"pd.set_option('display.max_colwidth', None)\n",
"display(pd.DataFrame(table))"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "16860fe8",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"Review | Label | Prediction (GPT-2) | Full prediction (GPT-2) | Prediction (GPT-Neo) | Full prediction (GPT-Neo) | Prediction (GPT-J) | Full prediction (GPT-J) | |
---|---|---|---|---|---|---|---|---|

0 | The best show I've seen so far this year! | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about the show. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

1 | Really enjoyed it | positive | positive | The reviewer is very happy about how the movie was made. The sentiment is positive. | positive | The reviewer enjoyed the movie. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

2 | Amazing. | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about how the movie was made. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

3 | Why all the hate? I enjoyed it. | positive | neutral | The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. | neutral | The reviewer is not sure why all the hate. The sentiment is neutral. | positive | The reviewer is happy about the movie. The sentiment is positive. |

4 | Beautiful visuals, entertaining, and I believe this show has a lot of potential! | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about the visuals, the entertainment, and the potential of the show. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

5 | A beautiful rendering of Middle Earth's history | positive | positive | The reviewer is happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about how the movie was made. The sentiment is positive. | positive | The reviewer is very happy about how the movie was made. The sentiment is positive. |

6 | So far, so good... and there's still hope | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very hopeful about the future of the movie. The sentiment is positive. | positive | The reviewer is optimistic about the movie. The sentiment is positive. |

7 | I'm a fan | positive | positive | The reviewer is a fan of the movie. The sentiment is positive. | positive | The reviewer is a fan of the movie. The sentiment is positive. | positive | The reviewer is a fan of the movie. The sentiment is positive. |

8 | It works for me | positive | positive | The reviewer is happy with the movie. The sentiment is positive. | positive | The reviewer likes the movie. The sentiment is positive. | positive | The reviewer is happy with the movie. The sentiment is positive. |

9 | Not the best, but enjoyed every episode. Can't wait to see much much more. | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | positive | The reviewer is happy about the movie. The sentiment is positive. |

10 | Beautiful, flawed, and a wonderful Fall treat | positive | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

11 | Good show with too many subplots | neutral | negative | The reviewer is not happy with the movie. The sentiment is negative. | positive | The reviewer is happy with the movie. The sentiment is positive. | positive | The reviewer is happy with the movie. The sentiment is positive. |

12 | Starts badly, gets better | neutral | negative | The reviewer is not happy with the movie. The sentiment is negative. | positive | The reviewer is not happy with the movie. The sentiment is positive. | neutral | The reviewer is not sure about the movie. The sentiment is neutral. |

13 | Good and bad things | neutral | positive | The reviewer is happy about the movie. The sentiment is positive. | positive | The reviewer is happy about the good things in the movie. The sentiment is positive. | positive | The reviewer is happy about the movie. The sentiment is positive. |

14 | Big and beautiful but can use a little help with its energy. | neutral | positive | The reviewer is happy about the movie. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. | positive | The reviewer is happy with the movie. The sentiment is positive. |

15 | Pretty but ultimately hollow and lacking in engagement | neutral | negative | The reviewer is not particularly happy with the movie. The sentiment is negative. | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | negative | The reviewer is disappointed by the movie. The sentiment is negative. |

16 | Beautiful visuals and story overshadowed by unnecessary gore and violence | neutral | negative | The reviewer is not happy with the movie. The sentiment is negative. | positive | The reviewer is very happy about the visuals and story. The sentiment is positive. | positive | The reviewer is very happy about the movie. The sentiment is positive. |

17 | Not what you're probably expecting | negative | positive | The reviewer is not sure what the movie is about. The sentiment is positive. | neutral | The reviewer is not what you are probably expecting. The sentiment is neutral. | negative | The reviewer is not happy with the movie. The sentiment is negative. |

18 | Poor writing; Uninteresting characters, nonsensical actions. | negative | neutral | The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. | negative | The reviewer is disappointed by the writing. The sentiment is negative. | negative | The reviewer is disappointed by the movie. The sentiment is negative. |

19 | Budget was spent on snacks between shots | negative | negative | The reviewer is not happy about the movie. The sentiment is negative. | positive | The reviewer is very happy about the budget. The sentiment is positive. | positive | The reviewer is happy about the movie. The sentiment is positive. |

20 | If you ignore the source material, it's still boring and weird | negative | positive | The reviewer is not sure about the movie. The sentiment is positive. | negative | The reviewer is bored by the source material. The sentiment is negative. | negative | The reviewer is not impressed by the movie. The sentiment is negative. |

21 | Just a bad show | negative | neutral | The reviewer does not feel particularly good or bad about the movie. The sentiment is neutral. | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | negative | The reviewer is disappointed by the movie. The sentiment is negative. |

22 | It's awful | negative | negative | The reviewer is very unhappy about the movie. The sentiment is negative. | negative | The reviewer is very disappointed by the movie. The sentiment is negative. | negative | The reviewer is very unhappy about the movie. The sentiment is negative. |

23 | I was hopeful... | negative | positive | The reviewer is hopeful about the movie. The sentiment is positive. | positive | The reviewer was hopeful about the movie. The sentiment is positive. | positive | The reviewer is hopeful about the movie. The sentiment is positive. |

24 | Painfully mediocre with a few good spots | negative | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | negative | The reviewer is not happy with the movie. The sentiment is negative. |

25 | Beautiful to look at... but that’s about it. | negative | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | positive | The reviewer is very happy about the movie. The sentiment is positive. | negative | The reviewer is not impressed by the movie. The sentiment is negative. |

26 | Underwhelming and disappointing | negative | negative | The reviewer is disappointed by the movie. The sentiment is negative. | negative | The reviewer is disappointed by the movie. The sentiment is negative. | negative | The reviewer is disappointed by the movie. The sentiment is negative. |

27 | Tolkien is rolling in his grave. No mystery. No inspiration. Wardrobe & acting is pretty bland. | negative | negative | The reviewer is not impressed by the movie. The sentiment is negative. | negative | The reviewer is not particularly happy about the movie. The sentiment is negative. | negative | The reviewer is not happy with the movie. The sentiment is negative. |