1_classifier_tfidf_svc.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "1_classifier_tfidf_svc.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/kshitijahande/Hate-Detection/blob/main/1_classifier_tfidf_svc.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZdAhZNPCeEm4"
      },
      "source": [
        "# Install libraries\n",
        "!pip install -qq ipynb\n",
        "!pip install -qq plotly\n",
        "!pip install -qq import-ipynb\n",
        "!pip install -qq icecream\n",
        "!pip install -qq transformers\n",
        "!pip install -qq datasets\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_NAfJkQTeKo3"
      },
      "source": [
        "#Import libraries\n",
        "import torch\n",
        "from transformers import AutoTokenizer, AutoModel\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import random as rn\n",
        "from icecream import ic \n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "from datasets import load_dataset\n",
        "from sklearn.model_selection import train_test_split\n",
        "import os\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aK23gAg_7myQ"
      },
      "source": [
        "# Load Dataset from Hugging Face library"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yoIVQmfPd_T9",
        "outputId": "ff6fe222-bb5e-4261-9731-e785fcffcaf3"
      },
      "source": [
        "dataset = load_dataset('hate_speech_offensive', split='train')\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using custom data configuration default\n",
            "Reusing dataset hate_speech_offensive (/root/.cache/huggingface/datasets/hate_speech_offensive/default/1.0.0/5f5dfc7b42b5c650fe30a8c49df90b7dbb9c7a4b3fe43ae2e66fabfea35113f5)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JlnZda7Yep2T",
        "outputId": "b5cc14ec-286b-484d-a50f-1ee7b30f7f02"
      },
      "source": [
        "#Columns in dataset\n",
        "ic(dataset[1]['tweet'])\n",
        "ic(dataset[1])\n",
        "ic(dataset.column_names)\n",
        "ic(dataset.num_rows)\n",
        "\n",
        "\"\"\"\n",
        "Task 1 - multiclass classification\n",
        "  \"class\":[\n",
        "              0:\"hate speech\"\n",
        "              1:\"offensive language\"\n",
        "              2:\"neither\"\n",
        "  ]\n",
        "\"\"\""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ic| dataset[1]['tweet']: ('!!!!! RT @mleew17: boy dats cold...tyga dwn bad for cuffin dat hoe in the '\n",
            "                          '1st place!!')\n",
            "ic| dataset[1]: {'class': 1,\n",
            "                 'count': 3,\n",
            "                 'hate_speech_count': 0,\n",
            "                 'neither_count': 0,\n",
            "                 'offensive_language_count': 3,\n",
            "                 'tweet': '!!!!! RT @mleew17: boy dats cold...tyga dwn bad for cuffin dat hoe '\n",
            "                          'in the 1st place!!'}\n",
            "ic| dataset.column_names: ['count',\n",
            "                           'hate_speech_count',\n",
            "                           'offensive_language_count',\n",
            "                           'neither_count',\n",
            "                           'class',\n",
            "                           'tweet']\n",
            "ic| dataset.num_rows: 24783\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "24783"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0PFqOQY2JM32",
        "outputId": "541629ea-9905-4989-957a-bdbc27a70d41"
      },
      "source": [
        "#Shuffle and split train and test dataset randomly\n",
        "train_test_split = dataset.train_test_split(shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Loading cached split indices for dataset at /root/.cache/huggingface/datasets/hate_speech_offensive/default/1.0.0/5f5dfc7b42b5c650fe30a8c49df90b7dbb9c7a4b3fe43ae2e66fabfea35113f5/cache-d21fa33e190556a1.arrow and /root/.cache/huggingface/datasets/hate_speech_offensive/default/1.0.0/5f5dfc7b42b5c650fe30a8c49df90b7dbb9c7a4b3fe43ae2e66fabfea35113f5/cache-0b5f0974b266240e.arrow\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vVc7Kw-tJmkP"
      },
      "source": [
        "#Get train and set data from original dataset\n",
        "test = train_test_split.get('test')\n",
        "train = train_test_split.get('train')\n",
        "\n",
        "#convert to data frame\n",
        "test_df = test.to_pandas()\n",
        "train_full_df = train.to_pandas()\n",
        "\n",
        "# split train dev/validation set\n",
        "train_df, val_df = train_test_split(train_full_df, test_size=0.2)\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CUsFUSoDVDUr",
        "outputId": "791d0462-efa8-451c-f0b9-21ca2bee9d7d"
      },
      "source": [
        "# Match total rows in each data frame\n",
        "ic(train_df.shape[0])\n",
        "ic(test_df.shape[0])\n",
        "ic(val_df.shape[0])\n",
        "total = train_df.shape[0] + test_df.shape[0] + val_df.shape[0]\n",
        "ic(total)\n",
        "ic(dataset.num_rows)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ic| train_df.shape[0]: 14869\n",
            "ic| test_df.shape[0]: 6196\n",
            "ic| dev_df.shape[0]: 3718\n",
            "ic| total: 24783\n",
            "ic| dataset.num_rows: 24783\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "24783"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "f_AiK4XKfRKh",
        "outputId": "02e00755-53dd-4a07-9f3f-82b176898532"
      },
      "source": [
        "# Hate vs non-hate distribution in data\n",
        "train_df['class'].value_counts(ascending=True)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1    11475\n",
              "2     2533\n",
              "0      861\n",
              "Name: class, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4KJSelosp0Ae"
      },
      "source": [
        "# FEATURE GENERATION"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gsKienV9p3kO",
        "outputId": "f63b9c48-8340-4863-f669-fd5649ddb8d1"
      },
      "source": [
        "from sklearn.feature_extraction.text import TfidfVectorizer\n",
        "import nltk\n",
        "from nltk.stem.porter import *\n",
        "import string\n",
        "import re\n",
        "\n",
        "nltk.download('stopwords')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/stopwords.zip.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 39
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_dPNT2BWlNFL"
      },
      "source": [
        "# Use tf-idf vectorizer\n",
        "\n",
        "other_exclusions = [\"#ff\", \"ff\", \"rt\"]\n",
        "stopwords=stopwords = nltk.corpus.stopwords.words(\"english\")\n",
        "stopwords.extend(other_exclusions)\n",
        "stemmer = PorterStemmer()\n",
        "\n",
        "def preprocess(text_string):\n",
        "    space_pattern = '\\s+'\n",
        "    giant_url_regex = ('http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|'\n",
        "        '[!*\\(\\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+')\n",
        "    mention_regex = '@[\\w\\-]+'\n",
        "    parsed_text = re.sub(space_pattern, ' ', text_string)\n",
        "    parsed_text = re.sub(giant_url_regex, '', parsed_text)\n",
        "    parsed_text = re.sub(mention_regex, '', parsed_text)\n",
        "    return parsed_text\n",
        "\n",
        "\"\"\"This function removes punctuation & excess whitespace, sets to lowercase,and stems tweets. Returns a list of stemmed tokens.\"\"\"\n",
        "\n",
        "def tokenize(tweet):\n",
        "    tweet = \" \".join(re.split(\"[^a-zA-Z]*\", tweet.lower())).strip()\n",
        "    #tokens = re.split(\"[^a-zA-Z]*\", tweet.lower())\n",
        "    tokens = [stemmer.stem(t) for t in tweet.split()]\n",
        "    return tokens\n",
        "\n",
        "vectorizer = TfidfVectorizer(\n",
        "    #vectorizer = sklearn.feature_extraction.text.CountVectorizer(\n",
        "    tokenizer=tokenize,\n",
        "    preprocessor=preprocess,\n",
        "    ngram_range=(1, 3),\n",
        "    stop_words=stopwords, #We do better when we keep stopwords\n",
        "    use_idf=True,\n",
        "    smooth_idf=False,\n",
        "    norm=None, #Applies l2 norm smoothing\n",
        "    decode_error='replace',\n",
        "    max_features=10000,\n",
        "    min_df=5,\n",
        "    max_df=0.501\n",
        "    )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eK87LXwXsUN9",
        "outputId": "c54f0657-7897-45b9-edab-a63f3279c2e1"
      },
      "source": [
        "# TFIDF matrix and scores\n",
        "tfidf = vectorizer.fit_transform(train_df['tweet']).toarray()\n",
        "vocab = {v:i for i, v in enumerate(vectorizer.get_feature_names())}\n",
        "idf_vals = vectorizer.idf_\n",
        "idf_dict = {i:idf_vals[i] for i in vocab.values()} #keys:indices; values:IDF scores"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/feature_extraction/text.py:385: UserWarning: Your stop_words may be inconsistent with your preprocessing. Tokenizing the stop words generated tokens ['b', 'c', 'e', 'f', 'g', 'h', 'j', 'l', 'n', 'p', 'r', 'u', 'v', 'w'] not in stop_words.\n",
            "  'stop_words.' % sorted(inconsistent))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8X9goUyOtgE6"
      },
      "source": [
        "# TASK 1 - Train and evaluate classifier"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4cmgaAjttmbP"
      },
      "source": [
        "X = pd.DataFrame(tfidf)\n",
        "y = train_df['class'].astype(int)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aBusvsBOt2PU"
      },
      "source": [
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.feature_selection import SelectFromModel\n",
        "from sklearn.metrics import classification_report\n",
        "from sklearn.svm import LinearSVC"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Zgmx5gQGt6EW",
        "outputId": "81000fd8-ca24-4af4-ce0e-1bab8d5e8a66"
      },
      "source": [
        "select = SelectFromModel(LogisticRegression(class_weight='balanced',penalty=\"l2\",C=0.01))\n",
        "X_ = select.fit_transform(X,y)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:940: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zxm47gubuzOe"
      },
      "source": [
        "model = LinearSVC(class_weight='balanced',C=0.01, penalty='l2', loss='squared_hinge',multi_class='ovr').fit(X_, y)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RX3HqMUhu5bM"
      },
      "source": [
        "y_preds = model.predict(X_)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ofUh0R5Ku9A-"
      },
      "source": [
        "report = classification_report( y, y_preds )\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "s_BQ12dhvAjt",
        "outputId": "ae4ac2b2-d349-4e3c-a930-83301d61f4ff"
      },
      "source": [
        "print(report)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.57      0.76      0.65       861\n",
            "           1       0.96      0.91      0.93     11475\n",
            "           2       0.78      0.90      0.84      2533\n",
            "\n",
            "    accuracy                           0.90     14869\n",
            "   macro avg       0.77      0.86      0.81     14869\n",
            "weighted avg       0.91      0.90      0.90     14869\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}