Lab04_Santiago_del_Rey_Juarez_Nikita_Belooussov.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Lab04-Santiago-del-Rey-Juarez_Nikita-Belooussov.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2jMkKylwoeys"
      },
      "source": [
        "Written by Santiagode Rey Juarez and Nikita Belooussov\n",
        "\n",
        "Consider Treebank corpus.\n",
        "\n",
        "Train HMM, TnT, perceptron and CRF models using the first 500, 1000, 1500, 2000, 2500 and 3000 sentences.\n",
        "\n",
        "Evaluate the resulting 24 models using sentences from 3001."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Y7QuqXTt5QIM",
        "outputId": "96202d07-d3ec-498c-e00e-9ed804693c62"
      },
      "source": [
        "!pip install python-crfsuite"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: python-crfsuite in /usr/local/lib/python3.7/dist-packages (0.9.7)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "f9ph04RSpuCe",
        "outputId": "6ae8eb41-08a4-4060-ed3f-33320edd624f"
      },
      "source": [
        "import nltk\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import time\n",
        "\n",
        "nltk.download('treebank')\n",
        "train_data1 = nltk.corpus.treebank.tagged_sents()[:500]\n",
        "train_data2 = nltk.corpus.treebank.tagged_sents()[:1000]\n",
        "train_data3 = nltk.corpus.treebank.tagged_sents()[:1500]\n",
        "train_data4 = nltk.corpus.treebank.tagged_sents()[:2000]\n",
        "train_data5 = nltk.corpus.treebank.tagged_sents()[:2500]\n",
        "train_data6 = nltk.corpus.treebank.tagged_sents()[:3000]\n",
        "dataArray=[train_data1,train_data2,train_data3,train_data4,train_data5,train_data6]\n",
        "test_data = nltk.corpus.treebank.tagged_sents()[3000:]\n",
        "\n",
        "x=[500,1000,1500,2000,2500,3000]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[nltk_data] Downloading package treebank to /root/nltk_data...\n",
            "[nltk_data]   Package treebank is already up-to-date!\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1I9m1g6KqJRJ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7bf149f4-cdff-4f3f-ecc0-9cea59dc19cc"
      },
      "source": [
        "#CRF\n",
        "CRF = nltk.tag.CRFTagger()\n",
        "CRF_results = []\n",
        "timeCRF=[]\n",
        "eval_time_CRF=[]\n",
        "for train_data in dataArray:\n",
        "  print(\"trainded on\", len(train_data))\n",
        "  start_time = time.time()\n",
        "  CRF.train(train_data,'crf_tagger_model')\n",
        "  time_spent = time.time() - start_time\n",
        "  timeCRF.append(time_spent)\n",
        "  start_time = time.time()\n",
        "  accuracy = round(CRF.evaluate(test_data), 3)\n",
        "  eval_time = time.time() - start_time\n",
        "  eval_time_CRF.append(eval_time)\n",
        "  CRF_results.append(accuracy)\n",
        "  print('accuracy: ' + str(accuracy))\n",
        "  test=CRF.tag(['the', 'men', 'attended', 'to', 'the', 'meetings'])\n",
        "  print(test)\n",
        "  print(\"--- trained in %s seconds ---\" % (time_spent))\n",
        "  print(\"--- evaluated in %s seconds ---\" % (eval_time))\n",
        "  print(\"\\n\\n\")\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "trainded on 500\n",
            "accuracy: 0.91\n",
            "[('the', 'DT'), ('men', 'NN'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 4.43401575088501 seconds ---\n",
            "--- evaluated in 1.0408265590667725 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 1000\n",
            "accuracy: 0.925\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 11.470499515533447 seconds ---\n",
            "--- evaluated in 0.8834707736968994 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 1500\n",
            "accuracy: 0.933\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 19.99679136276245 seconds ---\n",
            "--- evaluated in 0.907804012298584 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 2000\n",
            "accuracy: 0.939\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 32.10358715057373 seconds ---\n",
            "--- evaluated in 0.9139218330383301 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 2500\n",
            "accuracy: 0.943\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 45.309916734695435 seconds ---\n",
            "--- evaluated in 0.9121408462524414 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 3000\n",
            "accuracy: 0.947\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 54.9536292552948 seconds ---\n",
            "--- evaluated in 0.9338133335113525 seconds ---\n",
            "\n",
            "\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-28sxfUeoRFj",
        "outputId": "f27f47af-4af3-476b-feca-7e47e7b32d1e"
      },
      "source": [
        "#HMM example\n",
        "trainer = nltk.tag.hmm.HiddenMarkovModelTrainer()\n",
        "counter=1\n",
        "HMM_results = []\n",
        "timeHMM=[]\n",
        "eval_time_HMM = []\n",
        "for i in dataArray:\n",
        "    print(\"trained on \" + str(counter*500))\n",
        "    start_time = time.time()\n",
        "    HMM = trainer.train_supervised(i)\n",
        "    time_spent = time.time() - start_time\n",
        "    start_time = time.time()\n",
        "    accuracy = round(HMM.evaluate(test_data), 3)\n",
        "    eval_time = time.time() - start_time\n",
        "    eval_time_HMM.append(eval_time)\n",
        "    HMM_results.append(accuracy)\n",
        "    print('accuracy: ' + str(accuracy))\n",
        "    test=HMM.tag(['the', 'men', 'attended', 'to', 'the', 'meetings'])\n",
        "    print(test)\n",
        "    print(\"--- trained in %s seconds ---\" % (time_spent))\n",
        "    print(\"--- evaluated in %s seconds ---\" % (eval_time))\n",
        "    timeHMM.append(time_spent)\n",
        "    print(\"\\n\\n\")\n",
        "    counter=counter+1\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "trained on 500\n",
            "accuracy: 0.173\n",
            "[('the', 'DT'), ('men', 'NNP'), ('attended', 'NNP'), ('to', 'NNP'), ('the', 'NNP'), ('meetings', 'NNP')]\n",
            "--- trained in 0.20212769508361816 seconds ---\n",
            "--- evaluated in 8.579485177993774 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 1000\n",
            "accuracy: 0.221\n",
            "[('the', 'DT'), ('men', 'NNP'), ('attended', 'NNP'), ('to', 'NNP'), ('the', 'NNP'), ('meetings', 'NNP')]\n",
            "--- trained in 0.423386812210083 seconds ---\n",
            "--- evaluated in 9.376102209091187 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 1500\n",
            "accuracy: 0.262\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'NNP'), ('to', 'NNP'), ('the', 'NNP'), ('meetings', 'NNP')]\n",
            "--- trained in 0.6250088214874268 seconds ---\n",
            "--- evaluated in 10.018921136856079 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 2000\n",
            "accuracy: 0.304\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 0.8606200218200684 seconds ---\n",
            "--- evaluated in 10.307266235351562 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 2500\n",
            "accuracy: 0.335\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 1.0467722415924072 seconds ---\n",
            "--- evaluated in 10.499659299850464 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 3000\n",
            "accuracy: 0.368\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 1.2392351627349854 seconds ---\n",
            "--- evaluated in 10.600460290908813 seconds ---\n",
            "\n",
            "\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "t8KXVMBNpnYL",
        "outputId": "d9ee5c76-6d00-42e5-da9d-3e9473e934bf"
      },
      "source": [
        "#TnT example\n",
        "TnT = nltk.tag.tnt.TnT()\n",
        "TnT_results = []\n",
        "timeTnT=[]\n",
        "eval_time_TnT = []\n",
        "for train_data in dataArray:\n",
        "  print(\"trainded on\", len(train_data))\n",
        "  start_time = time.time()\n",
        "  TnT.train(train_data)\n",
        "  time_spent = time.time() - start_time\n",
        "  start_time = time.time()\n",
        "  accuracy = round(TnT.evaluate(test_data), 3)\n",
        "  eval_time = time.time() - start_time\n",
        "  eval_time_TnT.append(eval_time)\n",
        "  TnT_results.append(accuracy)\n",
        "  print('accuracy: ' + str(accuracy))\n",
        "  TnT.tag(['the', 'men', 'attended', 'to', 'the', 'meetings'])\n",
        "  print(test)\n",
        "  print(\"--- trained in %s seconds ---\" % (time_spent))\n",
        "  print(\"--- evaluated in %s seconds ---\" % (eval_time))\n",
        "  timeTnT.append(time_spent)\n",
        "  print(\"\\n\\n\")\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "trainded on 500\n",
            "accuracy: 0.747\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 0.3065512180328369 seconds ---\n",
            "--- evaluated in 17.582540273666382 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 1000\n",
            "accuracy: 0.796\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 0.5733785629272461 seconds ---\n",
            "--- evaluated in 45.03550720214844 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 1500\n",
            "accuracy: 0.827\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 0.8096468448638916 seconds ---\n",
            "--- evaluated in 73.73748540878296 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 2000\n",
            "accuracy: 0.848\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 1.076977252960205 seconds ---\n",
            "--- evaluated in 92.5980179309845 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 2500\n",
            "accuracy: 0.862\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 1.3325104713439941 seconds ---\n",
            "--- evaluated in 98.71757769584656 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trainded on 3000\n",
            "accuracy: 0.875\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 1.6038379669189453 seconds ---\n",
            "--- evaluated in 103.39435625076294 seconds ---\n",
            "\n",
            "\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JtcQG-GLp6Xj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6d8f6e72-e755-4350-ca37-fe60980f46df"
      },
      "source": [
        "#perceptron\n",
        "PER = nltk.tag.perceptron.PerceptronTagger(load=False)\n",
        "counter=1\n",
        "PER_results = []\n",
        "timePER=[]\n",
        "eval_time_PER = []\n",
        "for i in dataArray:\n",
        "    print(\"trained on \" +str(counter*500))\n",
        "    start_time = time.time()\n",
        "    PER.train(i)\n",
        "    time_spent = time.time() - start_time\n",
        "    start_time = time.time()\n",
        "    accuracy = round(PER.evaluate(test_data), 3)\n",
        "    eval_time = time.time() - start_time\n",
        "    eval_time_PER.append(eval_time)\n",
        "    PER_results.append(accuracy)\n",
        "    print('accuracy: ' + str(accuracy))\n",
        "    test=PER.tag(['the', 'men', 'attended', 'to', 'the', 'meetings'])\n",
        "    print(test)\n",
        "    print(\"--- trained in %s seconds ---\" % (time_spent))\n",
        "    print(\"--- evaluated in %s seconds ---\" % (eval_time))\n",
        "    timePER.append(time_spent)\n",
        "    print(\"\\n\\n\")\n",
        "    counter=counter+1\n",
        "\n",
        "# sns.lineplot(x=x, y=PER_results)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "trained on 500\n",
            "accuracy: 0.914\n",
            "[('the', 'DT'), ('men', 'NN'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 3.3177530765533447 seconds ---\n",
            "--- evaluated in 1.7758727073669434 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 1000\n",
            "accuracy: 0.932\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 6.700852155685425 seconds ---\n",
            "--- evaluated in 1.791055679321289 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 1500\n",
            "accuracy: 0.941\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 10.241178750991821 seconds ---\n",
            "--- evaluated in 1.8359756469726562 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 2000\n",
            "accuracy: 0.948\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 14.11820673942566 seconds ---\n",
            "--- evaluated in 1.8243582248687744 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 2500\n",
            "accuracy: 0.95\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 17.30457830429077 seconds ---\n",
            "--- evaluated in 1.8355369567871094 seconds ---\n",
            "\n",
            "\n",
            "\n",
            "trained on 3000\n",
            "accuracy: 0.953\n",
            "[('the', 'DT'), ('men', 'NNS'), ('attended', 'VBD'), ('to', 'TO'), ('the', 'DT'), ('meetings', 'NNS')]\n",
            "--- trained in 20.592247009277344 seconds ---\n",
            "--- evaluated in 1.812053918838501 seconds ---\n",
            "\n",
            "\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VY46DxSYoinj"
      },
      "source": [
        "Provide a figure with four learning curves, each per model type (X=training set size; Y=accuracy).\n",
        "\n",
        "Which model would you select? Justify the answer."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9e97HlwmoFNs"
      },
      "source": [
        "hmm_tag = ['HMM' for i in x]\n",
        "tnt_tag = ['TnT' for i in x]\n",
        "per_tag = ['PER' for i in x]\n",
        "crf_tag = ['CRF' for i in x]\n",
        "data = {'train size': [*x, *x, *x, *x],\n",
        "        'accuracy': [*HMM_results, *TnT_results, *PER_results, *CRF_results],\n",
        "        'model': [*hmm_tag, *tnt_tag, *per_tag, *crf_tag]}\n",
        "df_accuracy = pd.DataFrame(data)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uz3j2KkHDMoL"
      },
      "source": [
        "hmm_tag = ['HMM' for i in x]\n",
        "tnt_tag = ['TnT' for i in x]\n",
        "per_tag = ['PER' for i in x]\n",
        "crf_tag = ['CRF' for i in x]\n",
        "data = {'train size': [*x, *x, *x, *x],\n",
        "        'time': [*timeHMM, *timeTnT, *timePER, *timeCRF],\n",
        "        'model': [*hmm_tag, *tnt_tag, *per_tag, *crf_tag]}\n",
        "df_train_time = pd.DataFrame(data)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IOiCxZT1Qw3_"
      },
      "source": [
        "hmm_tag = ['HMM' for i in x]\n",
        "tnt_tag = ['TnT' for i in x]\n",
        "per_tag = ['PER' for i in x]\n",
        "crf_tag = ['CRF' for i in x]\n",
        "data = {'train size': [*x, *x, *x, *x],\n",
        "        'eval time': [*eval_time_HMM, *eval_time_TnT, *eval_time_PER, *eval_time_CRF],\n",
        "        'model': [*hmm_tag, *tnt_tag, *per_tag, *crf_tag]}\n",
        "df_eval_time = pd.DataFrame(data)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 404
        },
        "id": "r_O2yhR-U3Q-",
        "outputId": "7e70ec68-a22c-43ed-e1fc-415fed7fe48a"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.figure(figsize=(20,6))\n",
        "plt.subplot(131)\n",
        "sns.lineplot(data=df_accuracy, x='train size', y='accuracy', hue='model')\n",
        "plt.title(\"Models accuracy\")\n",
        "\n",
        "plt.subplot(132)\n",
        "sns.lineplot(data=df_train_time, x='train size', y='time', hue='model')\n",
        "plt.title(\"Models training time\")\n",
        "\n",
        "plt.subplot(133)\n",
        "sns.lineplot(data=df_eval_time, x='train size', y='eval time', hue='model')\n",
        "plt.title(\"Models evaluation time\")\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x432 with 3 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Tp7mKXELq7K"
      },
      "source": [
        "Looking at the results, the most accurate model was the PER model, which was closely followed by the CRF model. Between these two the PER model is better because it performs better as well as trains faster than the CRF model and has a similar evaluation time to the CRF model. The two fastest training models were TNT and HMM models. Between these two it may be better to use the TNT model, because it is vastly more accurate than the HMM model. A drawback to the TNT model is that it takes much longer to evaluate its results, compared to the HMM model. As a result, it depends mainly on what you are prioritizing, if it is speed in training then it is the TNT model, if it is accuracy or the speed of evaluation then it is the PER model."
      ]
    }
  ]
}