aravec-with-spacy.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "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": "3.7.4"
    },
    "colab": {
      "name": "aravec-with-spacy.ipynb",
      "provenance": [],
      "collapsed_sections": []
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N7nlMRmcJp7_",
        "colab_type": "text"
      },
      "source": [
        "## Integrate [AraVec](https://github.com/bakrianoo/aravec) with [Spacy.io](https://spacy.io/)\n",
        "\n",
        "This notebook demonstrates how to integrate an [AraVec](https://github.com/bakrianoo/aravec) model with [spaCy.io](https://spacy.io/)\n",
        "\n",
        "## Outlines\n",
        "\n",
        "- Install/Load the required modules\n",
        "- Load AraVec\n",
        "- Export the Word2Vec format + gzip it.\n",
        "- Initialize the spaCy model using AraVec vectors\n",
        "- Run Your AraVec Spacy Model\n",
        "- Test the Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aDX5kzUJJp8B",
        "colab_type": "text"
      },
      "source": [
        "## Install/Load the required modules"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-27T22:45:44.689719Z",
          "start_time": "2020-04-27T22:45:42.787139Z"
        },
        "scrolled": true,
        "id": "qrruGV8IJp8C",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install gensim spacy nltk"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jr-fEfMknBZH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import gensim\n",
        "import re\n",
        "import spacy\n",
        "\n",
        "# Clean/Normalize Arabic Text\n",
        "def clean_str(text):\n",
        "    search = [\"أ\",\"إ\",\"آ\",\"ة\",\"_\",\"-\",\"/\",\".\",\"،\",\" و \",\" يا \",'\"',\"ـ\",\"'\",\"ى\",\"\\\\\",'\\n', '\\t','&quot;','?','؟','!']\n",
        "    replace = [\"ا\",\"ا\",\"ا\",\"ه\",\" \",\" \",\"\",\"\",\"\",\" و\",\" يا\",\"\",\"\",\"\",\"ي\",\"\",' ', ' ',' ',' ? ',' ؟ ',' ! ']\n",
        "    \n",
        "    #remove tashkeel\n",
        "    p_tashkeel = re.compile(r'[\\u0617-\\u061A\\u064B-\\u0652]')\n",
        "    text = re.sub(p_tashkeel,\"\", text)\n",
        "    \n",
        "    #remove longation\n",
        "    p_longation = re.compile(r'(.)\\1+')\n",
        "    subst = r\"\\1\\1\"\n",
        "    text = re.sub(p_longation, subst, text)\n",
        "    \n",
        "    text = text.replace('وو', 'و')\n",
        "    text = text.replace('يي', 'ي')\n",
        "    text = text.replace('اا', 'ا')\n",
        "    \n",
        "    for i in range(0, len(search)):\n",
        "        text = text.replace(search[i], replace[i])\n",
        "    \n",
        "    #trim    \n",
        "    text = text.strip()\n",
        "\n",
        "    return text"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8_-qAgYQJp8L",
        "colab_type": "text"
      },
      "source": [
        "## Load AraVec\n",
        "Download a model from the [AraVec Repository](https://github.com/bakrianoo/aravec), then follow the below steps to load it."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T5a8VHhbij2f",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Download via terminal commands\n",
        "!wget \"https://bakrianoo.sfo2.digitaloceanspaces.com/aravec/full_grams_cbow_100_twitter.zip\"\n",
        "!unzip \"full_grams_cbow_100_twitter.zip\""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mjs_I0TUjGMP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "978ffd5e-5512-4931-d0ec-6c1a86fb01d2"
      },
      "source": [
        "# load the AraVec model\n",
        "model = gensim.models.Word2Vec.load(\"full_grams_cbow_100_twitter.mdl\")\n",
        "print(\"We've\",len(model.wv.index2word),\"vocabularies\")"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "We've 1476715 vocabularies\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ll25Yq9KJp8d",
        "colab_type": "text"
      },
      "source": [
        "## Export the Word2Vec format + gzip it."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-28T13:04:35.872005Z",
          "start_time": "2020-04-28T13:04:35.750353Z"
        },
        "id": "eYO6GPSSJp8e",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# make a directory called \"spacyModel\"\n",
        "%mkdir spacyModel"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-28T13:05:07.649631Z",
          "start_time": "2020-04-28T13:04:37.108481Z"
        },
        "id": "qrISD4ONJp8g",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# export the word2vec fomart to the directory\n",
        "model.wv.save_word2vec_format(\"./spacyModel/aravec.txt\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-28T13:05:47.899536Z",
          "start_time": "2020-04-28T13:05:07.650778Z"
        },
        "id": "BvpYpLXHJp8k",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# using `gzip` to compress the .txt file\n",
        "!gzip ./spacyModel/aravec.txt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mIEFJtxSJp8n",
        "colab_type": "text"
      },
      "source": [
        "## Initialize the spaCy model using AraVec vectors\n",
        "\n",
        "- This will create a folder called `/spacy.aravec.model` within your current working directory.\n",
        "- This step could take several minutes to be completed."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-28T13:06:08.570498Z",
          "start_time": "2020-04-28T13:05:49.692840Z"
        },
        "id": "klLd5QoeJp8o",
        "colab_type": "code",
        "outputId": "95ec1a6d-1a6d-4edd-a464-960ed014db3e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 101
        }
      },
      "source": [
        "!python -m spacy  init-model ar spacy.aravec.model --vectors-loc ./spacyModel/aravec.txt.gz"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[2K\u001b[38;5;2m✔ Successfully created model\u001b[0m\n",
            "1476715it [01:04, 22923.32it/s]\n",
            "\u001b[2K\u001b[38;5;2m✔ Loaded vectors from spacyModel/aravec.txt.gz\u001b[0m\n",
            "\u001b[38;5;2m✔ Sucessfully compiled vocab\u001b[0m\n",
            "1476903 entries, 1476715 vectors\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ReikgJ93Jp8r",
        "colab_type": "text"
      },
      "source": [
        "## Run Your AraVec Spacy Model\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2020-04-28T13:06:25.120513Z",
          "start_time": "2020-04-28T13:06:24.032773Z"
        },
        "id": "ab-Hnj4kJp8r",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# load AraVec Spacy model\n",
        "nlp = spacy.load(\"./spacy.aravec.model/\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1Efk9S9onT4V",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Define the preprocessing Class\n",
        "class Preprocessor:\n",
        "    def __init__(self, tokenizer, **cfg):\n",
        "        self.tokenizer = tokenizer\n",
        "\n",
        "    def __call__(self, text):\n",
        "        preprocessed = clean_str(text)\n",
        "        return self.tokenizer(preprocessed)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9CIWoqhInihz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Apply the `Preprocessor` Class\n",
        "nlp.tokenizer = Preprocessor(nlp.tokenizer)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zYzrWSHLn0sM",
        "colab_type": "text"
      },
      "source": [
        "## Test the Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "X8QZEGiunuUT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 370
        },
        "outputId": "fdc0f830-c37b-4323-ba2c-f102d2c15fe9"
      },
      "source": [
        "# Test your model\n",
        "nlp(\"قطة\").vector"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([ 0.6214019 ,  2.664876  , -2.4490244 , -0.13141291,  1.0106287 ,\n",
              "        1.4277642 , -0.6019407 , -0.37155798,  2.2610269 , -0.51503485,\n",
              "       -1.7400011 ,  1.4599515 ,  1.3110927 ,  0.4506139 ,  1.1511235 ,\n",
              "       -2.3989084 ,  0.0108205 , -0.93597263,  0.20742278,  2.7626824 ,\n",
              "       -0.21789424, -2.6269352 , -0.033042  , -2.0458148 ,  1.4766251 ,\n",
              "       -2.589866  , -1.7341375 , -1.5589778 ,  0.57571614,  4.2727513 ,\n",
              "        0.02701492,  1.77316   , -1.1816478 ,  0.24516247, -0.04227808,\n",
              "        0.57215565,  3.2628767 , -1.2422727 ,  1.2351261 ,  1.7213373 ,\n",
              "       -2.2107098 ,  3.334359  ,  0.835815  ,  0.27691752, -0.61994714,\n",
              "        2.0607152 , -0.33151346,  2.132865  , -1.1516991 , -0.39679298,\n",
              "       -2.1682317 ,  1.5982645 , -1.1571178 ,  1.3672193 , -0.81996626,\n",
              "        0.5634883 ,  0.8571397 ,  1.2602032 ,  1.5811064 , -2.6346667 ,\n",
              "       -0.21950944, -1.7665412 ,  1.3162723 , -0.9176698 , -0.5075662 ,\n",
              "       -0.6396452 , -0.57308793,  2.6602883 ,  1.466169  , -0.54523975,\n",
              "        1.0440696 , -0.8016639 ,  0.95874494, -0.8008114 ,  1.4913949 ,\n",
              "        0.9796351 , -1.3504812 , -0.16031194,  0.779816  , -2.036837  ,\n",
              "        0.5117812 , -1.174033  ,  3.356553  ,  1.5459414 , -1.1024675 ,\n",
              "        1.4124179 ,  1.0076581 , -0.23065878,  3.9290988 ,  0.24867593,\n",
              "       -0.8912038 ,  0.7108348 ,  0.40351257,  3.1929119 , -0.7811022 ,\n",
              "       -2.341077  ,  0.38009226, -0.7102923 , -0.6132934 , -0.88354295],\n",
              "      dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q25t-eoAn0Kj",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 50
        },
        "outputId": "01491f6f-8739-4d07-8dc6-a3b3b2c9be23"
      },
      "source": [
        "egypt = nlp(\"مصر\")\n",
        "tunisia = nlp(\"تونس\")\n",
        "apple = nlp(\"تفاح\")\n",
        "\n",
        "print(\"egypt Vs. tunisia = \", egypt.similarity(tunisia))\n",
        "print(\"egypt Vs. apple = \", egypt.similarity(apple))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "egypt Vs. tunisia =  0.8277813628455393\n",
            "egypt Vs. apple =  0.09689554003631644\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4iFhK4f-Jp9G",
        "colab_type": "text"
      },
      "source": [
        "## Done !!\n",
        "\n",
        "Congratulations, now you have your AraVec model running on spaCy."
      ]
    }
  ]
}