Semantic_Search_BERT.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Semantic_Search_BERT",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOuWehZMfpDm+EpzIgW218v",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ludwigwittgenstein2/Research/blob/master/Semantic_Search_BERT.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LOAZQoLqj6XR"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WhTYsz81j-LX"
      },
      "source": [
        "# Semantic Search with BERT \n",
        "  by Rick Rejeleene\n",
        "\n",
        "```\n",
        "` This is formatted as code`\n",
        "\n",
        "Objective: Apply Query: \"I like Fruits.\" \n",
        "            and rank results with probabilities\n",
        "\n",
        "```\n",
        "#0. Can we apply BERT to this problem? \n",
        "  To find Similarity between two sentences,\n",
        "  We can feed each unique pair through BERT, \n",
        "  and compare it to all scores. It would take for 10 sentences, 49 passes through BERT, which a GPU would take 60+ hours.\n",
        "\n",
        "\n",
        "  So, instead we pre compute representation for each sequence in our dataset.\n",
        "\n",
        "  Averaging the word embedding for sentence embedding provides worse than GloVE embedding\n",
        "\n",
        "# 1. What is Sentence Embedding? \n",
        "\n",
        "Text Embeddings are vector representation of words. \n",
        "\n",
        "Sentence Embedding are vector representation of sentences. \n",
        "\n",
        "\n",
        "#2.What is Semantic Search?\n",
        "Semantic Search is search with meaning, in contrast with lexical search. \n",
        "\n",
        "In lexical search, our search matches literal query or variants of it. \n",
        "\n",
        "Semantic Search on the other hand, searches with context\n",
        "\n",
        "Eg: \"I like apples.\" from corupus of 10 sentences would bring up most probable matches and rank sentences. \n",
        "\n",
        "#3.How do we use Semantic Search with BERT? \n",
        "\n",
        "We use SBERT. \n",
        "\n",
        "In SBERT, we process two sentences at the same time using Siamese Network\n",
        "\n",
        "Siamese Network, in simple terms, is a neural network that specializes in measuring similarity, widely used in Facial Recognition systems. \n",
        "\n",
        "We have BERT as the base of this model, and we add Pooling Layer to represent input sentences of various length. \n",
        "\n",
        "U and V are sentence embedding, that are concantenated before multiplication. \n",
        "\n",
        "Softmax gives us probability of each class\n",
        "\n",
        "#4. How does this work? Give me the Architecture of this\n",
        "\n",
        "![1_GNhALCfeEGz5JaXWjc106w.png](data:image/png;base64,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)\n",
        "\n",
        "1. Sentence A and Sentence B are passed simultaneously \n",
        "2. BERT takes in both sentences\n",
        "3. We add Pooling layer, that creates fixed vector represenation of input sentences of varying length\n",
        "4. U and V are sentence embedding\n",
        "5. U and V are concantented\n",
        "6. U and V are multiplied \n",
        "7. We sent it to SoftMax Function, which is a generalization of logistic function\n",
        "\n",
        "\n",
        "\n",
        "#Source: \n",
        "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a0aWzPGalRTx",
        "outputId": "d48c452f-e1a2-4334-8386-0db6c095b0ac",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "#Install Library called Sentence Transformer from Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks\n",
        "\n",
        "!pip install sentence-transformers"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting sentence-transformers\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/f4/fd/0190080aa0af78d7cd5874e4e8e85f0bed9967dd387cf05d760832b95da9/sentence-transformers-0.3.8.tar.gz (66kB)\n",
            "\u001b[K     |████████████████████████████████| 71kB 3.3MB/s \n",
            "\u001b[?25hCollecting transformers<3.4.0,>=3.1.0\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/19/22/aff234f4a841f8999e68a7a94bdd4b60b4cebcfeca5d67d61cd08c9179de/transformers-3.3.1-py3-none-any.whl (1.1MB)\n",
            "\u001b[K     |████████████████████████████████| 1.1MB 3.6MB/s \n",
            "\u001b[?25hRequirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (4.41.1)\n",
            "Requirement already satisfied: torch>=1.2.0 in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (1.7.0+cu101)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (1.18.5)\n",
            "Requirement already satisfied: scikit-learn in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (0.22.2.post1)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (1.4.1)\n",
            "Requirement already satisfied: nltk in /usr/local/lib/python3.6/dist-packages (from sentence-transformers) (3.2.5)\n",
            "Collecting tokenizers==0.8.1.rc2\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/80/83/8b9fccb9e48eeb575ee19179e2bdde0ee9a1904f97de5f02d19016b8804f/tokenizers-0.8.1rc2-cp36-cp36m-manylinux1_x86_64.whl (3.0MB)\n",
            "\u001b[K     |████████████████████████████████| 3.0MB 13.6MB/s \n",
            "\u001b[?25hCollecting sacremoses\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7d/34/09d19aff26edcc8eb2a01bed8e98f13a1537005d31e95233fd48216eed10/sacremoses-0.0.43.tar.gz (883kB)\n",
            "\u001b[K     |████████████████████████████████| 890kB 35.5MB/s \n",
            "\u001b[?25hCollecting sentencepiece!=0.1.92\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e5/2d/6d4ca4bef9a67070fa1cac508606328329152b1df10bdf31fb6e4e727894/sentencepiece-0.1.94-cp36-cp36m-manylinux2014_x86_64.whl (1.1MB)\n",
            "\u001b[K     |████████████████████████████████| 1.1MB 43.9MB/s \n",
            "\u001b[?25hRequirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from transformers<3.4.0,>=3.1.0->sentence-transformers) (2.23.0)\n",
            "Requirement already satisfied: dataclasses; python_version < \"3.7\" in /usr/local/lib/python3.6/dist-packages (from transformers<3.4.0,>=3.1.0->sentence-transformers) (0.7)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from transformers<3.4.0,>=3.1.0->sentence-transformers) (20.4)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.6/dist-packages (from transformers<3.4.0,>=3.1.0->sentence-transformers) (3.0.12)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.6/dist-packages (from transformers<3.4.0,>=3.1.0->sentence-transformers) (2019.12.20)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.6/dist-packages (from torch>=1.2.0->sentence-transformers) (3.7.4.3)\n",
            "Requirement already satisfied: future in /usr/local/lib/python3.6/dist-packages (from torch>=1.2.0->sentence-transformers) (0.16.0)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn->sentence-transformers) (0.17.0)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from nltk->sentence-transformers) (1.15.0)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers<3.4.0,>=3.1.0->sentence-transformers) (7.1.2)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<3.4.0,>=3.1.0->sentence-transformers) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<3.4.0,>=3.1.0->sentence-transformers) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<3.4.0,>=3.1.0->sentence-transformers) (2020.6.20)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<3.4.0,>=3.1.0->sentence-transformers) (1.24.3)\n",
            "Requirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from packaging->transformers<3.4.0,>=3.1.0->sentence-transformers) (2.4.7)\n",
            "Building wheels for collected packages: sentence-transformers, sacremoses\n",
            "  Building wheel for sentence-transformers (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for sentence-transformers: filename=sentence_transformers-0.3.8-cp36-none-any.whl size=101996 sha256=c27123a5a8e5da49cb8486eff9465dfb91517348a033d2214547bdcd6d026e51\n",
            "  Stored in directory: /root/.cache/pip/wheels/27/ec/b3/d12cc8e4daf77846db6543033d3a5642f204c0320b15945647\n",
            "  Building wheel for sacremoses (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for sacremoses: filename=sacremoses-0.0.43-cp36-none-any.whl size=893257 sha256=1c804ba3653839d9a6ea62aa3c2d5dd5deed711a7497612165daa94cca4bb557\n",
            "  Stored in directory: /root/.cache/pip/wheels/29/3c/fd/7ce5c3f0666dab31a50123635e6fb5e19ceb42ce38d4e58f45\n",
            "Successfully built sentence-transformers sacremoses\n",
            "Installing collected packages: tokenizers, sacremoses, sentencepiece, transformers, sentence-transformers\n",
            "Successfully installed sacremoses-0.0.43 sentence-transformers-0.3.8 sentencepiece-0.1.94 tokenizers-0.8.1rc2 transformers-3.3.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xe0MF7Ffnzcv"
      },
      "source": [
        "#Location of Library for Documentation\n",
        "#https://pypi.org/project/sentence-transformers/\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NnJGOKyMlVNb",
        "outputId": "8b7a8b6e-3f43-491e-c4a3-043aaa32f548",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "#Import Library, and import module Sentence Transformer\n",
        "from sentence_transformers import SentenceTransformer\n",
        "\n",
        "#We based it bert-base-nli-mean-tokens model from the sentence-transformers-repository.\n",
        "model = SentenceTransformer('bert-base-nli-mean-tokens')\n",
        "\n",
        "#We give sample sentences\n",
        "sentences = ['Who won the election?', 'Lack of correct news source', 'How do we figure out who won the election?', 'Did everyone win in the election?']\n",
        "\n",
        "#Encode Sentences\n",
        "sentence_embeddings = model.encode(sentences)\n",
        "\n",
        "#Print Embedding\n",
        "print('SAMPLE BERT embedding Vector-length', len(sentence_embeddings[0]))\n",
        "\n",
        "print('Sample BERT embedding vector - note includes negative values', sentence_embeddings[0])\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "SAMPLE BERT embedding Vector-length 768\n",
            "Sample BERT embedding vector - note includes negative values [-5.06621242e-01 -4.14385647e-01  6.80156589e-01  1.10902905e+00\n",
            "  1.38112828e-01  7.03566074e-01 -3.20728481e-01  4.63718623e-01\n",
            " -5.77143252e-01  4.42124158e-01  6.48557961e-01  1.66368354e-02\n",
            "  8.72302830e-01 -1.19855806e-01 -4.56467196e-02  2.80204505e-01\n",
            " -3.70686293e-01 -2.12453958e-02  4.91884083e-01 -4.65284914e-01\n",
            "  1.25715330e-01  7.16980517e-01 -9.05266821e-01 -1.36717096e-01\n",
            "  2.06467032e-01  4.85025644e-01  3.01887784e-02 -1.98085093e+00\n",
            " -9.38615978e-01 -2.74710894e-01 -5.94572604e-01  1.81638345e-01\n",
            " -5.83316207e-01 -2.96293199e-01 -7.51849592e-01  3.89402539e-01\n",
            "  9.63532150e-01 -4.41591322e-01 -3.46491158e-01  3.63579452e-01\n",
            " -2.16554686e-01 -7.27016091e-01  1.25904596e+00  7.77477622e-01\n",
            " -8.05147648e-01 -5.38869560e-01 -6.91540062e-01  9.94445324e-01\n",
            "  5.82817435e-01 -1.14126527e+00 -8.17383647e-01 -1.21538436e+00\n",
            "  8.52550268e-02  1.29542649e+00 -3.38971704e-01  5.49694836e-01\n",
            "  7.81894028e-01 -8.00547957e-01 -7.97830939e-01 -2.10113689e-01\n",
            " -7.13036358e-01  7.58189633e-02  4.17631269e-02 -3.51696253e-01\n",
            "  5.62554896e-01  3.28210741e-01  1.09872305e+00  5.38778663e-01\n",
            " -7.92275131e-01 -2.78749287e-01  2.13550344e-01 -4.65425909e-01\n",
            " -6.00414455e-01  5.62883735e-01 -8.79303634e-01 -2.47035533e-01\n",
            " -2.52361268e-01  1.21943021e+00  6.03757262e-01 -1.36758432e-01\n",
            " -2.37430766e-01 -1.71645418e-01  2.51564443e-01  9.16983962e-01\n",
            "  5.59478402e-02  7.95141876e-01  2.24310234e-01 -1.90529108e-01\n",
            " -6.13068104e-01  1.82478756e-01 -4.43221956e-01 -6.36259496e-01\n",
            "  6.93977237e-01  3.16597909e-01 -4.77979511e-01 -1.45938978e-01\n",
            "  9.22567964e-01 -5.29865384e-01  3.57058868e-02 -6.26335263e-01\n",
            "  2.26461500e-01 -2.53765970e-01  8.75026956e-02  1.20884962e-01\n",
            " -4.51414555e-01  1.91887289e-01 -3.41706038e-01 -8.35058272e-01\n",
            "  8.11110318e-01  1.08987677e+00  4.89812225e-01  1.22794561e-01\n",
            "  9.68999684e-01  2.49377832e-01 -8.82793963e-01  3.03792447e-01\n",
            " -3.39336663e-01 -1.86138704e-01  4.60406512e-01  1.35373986e+00\n",
            "  5.81561983e-01  7.19952762e-01 -9.99105990e-01  1.64881974e-01\n",
            " -1.03905328e-01  1.55906036e-01 -3.32516700e-01  1.04284120e+00\n",
            " -8.31222236e-01 -2.70580381e-01 -3.57692279e-02  5.76033235e-01\n",
            "  5.75945489e-02 -1.96410030e-01  4.83020283e-02  2.00668305e-01\n",
            " -1.56923339e-01 -1.09625801e-01 -3.90208691e-01 -1.23370029e-01\n",
            " -1.81747302e-01  1.98548079e-01  6.80457473e-01 -1.83123037e-01\n",
            " -7.77893186e-01 -2.78302610e-01  9.80172694e-01 -1.12220824e-01\n",
            " -1.14598684e-01 -4.51101035e-01  1.31832376e-01  6.12174213e-01\n",
            " -1.22080646e-01 -9.34540451e-01 -7.06749499e-01  3.96566212e-01\n",
            "  1.23852573e-01 -5.81680954e-01  5.83052218e-01  8.19851995e-01\n",
            "  3.27097267e-01 -2.20227554e-01  9.74699974e-01  6.37135208e-01\n",
            "  4.70363557e-01 -3.83418292e-01  2.16573167e+00  1.05238628e+00\n",
            "  3.03573638e-01 -1.42209083e-01 -3.50230522e-02 -1.26893684e-01\n",
            "  1.16361201e+00  1.07044435e+00  1.27231812e+00  3.03830832e-01\n",
            "  1.59678757e-02  4.40307766e-01  5.23983479e-01  1.38387317e-02\n",
            " -1.06247865e-01 -6.50011063e-01 -3.54584932e-01  1.92173600e-01\n",
            "  2.35713795e-01  2.36958310e-01 -1.30411789e-01 -4.58209306e-01\n",
            "  6.71921790e-01  9.36375633e-02 -4.58221108e-01 -4.79137689e-01\n",
            " -1.87046796e-01 -1.06227541e+00 -2.54243463e-02  3.56981546e-01\n",
            "  5.74348330e-01  9.33067679e-01  5.12960181e-03 -3.78626078e-01\n",
            "  5.45139611e-01 -6.34074509e-01  4.58790928e-01 -5.33372462e-01\n",
            "  4.42922175e-01  7.08641827e-01  7.75974631e-01  7.91530430e-01\n",
            " -2.23037153e-01 -6.64775074e-01 -9.60326314e-01  5.66918373e-01\n",
            "  1.03287196e+00  3.69739324e-01 -1.69054672e-01 -3.18551689e-01\n",
            "  1.16611004e-01  3.02994519e-01  1.01985872e+00  4.36940759e-01\n",
            "  1.17578948e+00  6.52212575e-02  9.59027827e-01  3.36042672e-01\n",
            " -1.42483830e-01  1.27246571e+00  2.41251871e-01 -1.00279009e+00\n",
            " -9.83150482e-01  7.65941441e-01  7.60373771e-01  7.30393305e-02\n",
            " -2.45392218e-01 -1.78172573e-01 -1.06529899e-01 -6.61946714e-01\n",
            " -1.51972699e+00 -3.80910248e-01  9.86087918e-01  1.19296931e-01\n",
            "  8.03503573e-01  3.23269427e-01  1.05298948e+00  1.14548051e+00\n",
            " -6.42937720e-01  1.29693955e-01  7.45421410e-01 -1.21303901e-01\n",
            "  8.43643546e-02  4.50865716e-01 -2.47300729e-01 -1.17977834e+00\n",
            " -5.18259168e-01 -6.24969304e-01 -1.49675846e-01  2.26975769e-01\n",
            "  2.45415971e-01  4.70207691e-01  2.75608838e-01 -8.76161531e-02\n",
            "  1.80650234e-01  4.32815440e-02 -4.58632052e-01  4.16567832e-01\n",
            "  2.58805305e-01 -4.99187142e-01 -7.85007179e-01 -1.21944018e-01\n",
            " -7.09517360e-01  6.36801496e-02 -4.43731636e-01  6.35920791e-03\n",
            "  7.56718740e-02  5.74637294e-01  5.60437083e-01 -6.42821968e-01\n",
            "  5.94961531e-02 -3.24091613e-02 -9.44177330e-01 -9.63406563e-02\n",
            " -7.69446671e-01 -4.42013264e-01 -2.41039157e+00 -1.47993967e-01\n",
            " -2.00406402e-01  6.44444823e-01  4.39249188e-01 -3.70947629e-01\n",
            " -2.44675539e-02 -1.12496905e-01 -4.96610582e-01 -2.86785781e-01\n",
            "  2.42465287e-01 -1.28779626e+00  1.03572023e+00 -5.05205154e-01\n",
            "  7.59102941e-01 -1.11322534e+00 -4.17147398e-01  4.66815948e-01\n",
            "  4.50095609e-02  1.84063509e-01 -5.72663307e-01 -2.06060410e-01\n",
            "  6.84923353e-03 -7.02888668e-01  4.90236759e-01 -9.11040828e-02\n",
            " -1.85852796e-01  1.67484403e-01 -2.57497102e-01 -1.53168678e-01\n",
            "  5.31130493e-01  3.10073555e-01 -4.41316426e-01  6.99916720e-01\n",
            " -5.20438612e-01 -2.86447823e-01  8.28841031e-01 -7.24966347e-01\n",
            " -6.56537890e-01  4.38847870e-01 -1.61705464e-01 -3.43494833e-01\n",
            "  7.28282094e-01 -6.80726394e-02 -7.99839377e-01  5.34885466e-01\n",
            " -3.64640862e-01 -2.30137497e-01  1.31908059e+00 -1.79621428e-01\n",
            " -2.91404545e-01  1.40649402e+00  5.96317708e-01  2.07314700e-01\n",
            "  6.77120090e-01 -3.04842502e-01 -2.68078712e-03  1.02204844e-01\n",
            "  1.37786508e+00  5.48654318e-01 -4.32411492e-01 -7.88049400e-03\n",
            "  7.80460298e-01 -1.72541454e-01 -4.91871417e-01 -5.85571565e-02\n",
            "  2.20296368e-01  2.44248852e-01  3.46702337e-01 -1.12335697e-01\n",
            "  1.06022620e+00 -6.34767473e-01 -4.16495621e-01  5.71760774e-01\n",
            " -1.00414681e+00 -4.69556183e-01 -1.05269551e-01  9.55871761e-01\n",
            "  3.26182038e-01  6.60006166e-01 -8.33429158e-01  1.84207588e-01\n",
            " -1.34210214e-01 -6.13249302e-01  4.97522354e-01  7.85711184e-02\n",
            "  8.54736328e-01  8.45410168e-01  4.75908756e-01 -5.52886128e-02\n",
            " -2.65911907e-01 -5.44122577e-01 -2.57746540e-02 -6.23204075e-02\n",
            "  5.49027205e-01 -1.08345218e-01 -4.43934977e-01 -7.71154761e-01\n",
            "  1.02503367e-01 -5.84799290e-01 -3.15862685e-01 -2.55733877e-01\n",
            "  8.48105311e-01 -5.76008372e-02 -2.63972819e-01  2.48741627e-01\n",
            "  2.95493275e-01 -1.04883060e-01  9.65982854e-01  1.52779311e-01\n",
            "  2.76873559e-01  4.30280000e-01 -1.89891681e-01 -4.68182981e-01\n",
            "  6.60078347e-01  3.15506160e-01  3.31001543e-02 -1.51148006e-01\n",
            "  2.26412252e-01  3.64467353e-01  6.37980163e-01  3.71866673e-01\n",
            "  9.12409872e-02 -6.82732165e-01  7.29354084e-01  1.92246839e-01\n",
            " -1.29534258e-02 -2.73072034e-01 -3.69735718e-01 -2.44318083e-01\n",
            "  1.49103984e-01  1.07916422e-01 -1.14618933e+00  3.86900127e-01\n",
            " -1.79330871e-01 -1.45404339e+00  2.00315088e-01 -7.92116046e-01\n",
            " -7.36707002e-02 -7.10118413e-02 -5.08834898e-01 -6.98574126e-01\n",
            " -1.59348816e-01  5.10965407e-01  4.48304594e-01  4.31337655e-02\n",
            "  1.28224880e-01  5.21345258e-01 -2.12861523e-01 -1.63844213e-01\n",
            " -2.41834903e+00 -1.27441054e-02 -6.17736399e-01 -7.18414187e-01\n",
            "  1.14521468e+00  5.26754737e-01  7.17795789e-01 -9.01705265e-01\n",
            " -3.93570423e-01 -3.10594559e-01 -5.31367779e-01  4.07052755e-01\n",
            "  1.29768765e+00 -2.52368182e-01 -2.87563026e-01 -2.26230845e-01\n",
            "  1.58758163e-01 -3.66150886e-01  9.61833000e-01 -1.08231235e+00\n",
            " -1.04434475e-01 -7.90580213e-01 -4.47804064e-01  4.15516227e-01\n",
            " -4.02394712e-01  2.71277819e-02 -3.61017406e-01 -7.54709184e-01\n",
            "  1.93078890e-01  1.32614523e-01 -6.81745112e-01 -5.52681983e-01\n",
            "  6.61115587e-01  3.75429571e-01 -6.16273522e-01 -1.22940195e+00\n",
            "  6.09291755e-02  2.85260119e-02  1.35868624e-01  1.33911952e-01\n",
            "  2.55617321e-01 -2.42183939e-01  4.84182648e-02 -4.54447508e-01\n",
            " -7.49498725e-01  1.09697890e+00  4.08254534e-01 -2.60929495e-01\n",
            "  1.87142536e-01 -1.72068954e-01 -6.47770762e-02 -5.03069341e-01\n",
            " -3.25681455e-02  5.56416214e-01 -5.11250019e-01 -1.01526417e-01\n",
            "  3.77250522e-01 -1.94040537e-01 -2.69037068e-01 -6.85654998e-01\n",
            " -4.69733179e-01 -9.32026446e-01 -3.51196975e-01 -8.30109790e-02\n",
            " -1.13111889e+00 -3.32525194e-01  1.39867768e-01  6.12420961e-02\n",
            "  4.77420278e-02 -8.19174349e-01 -6.91160142e-01 -1.03611402e-01\n",
            "  4.30121571e-01 -8.11179101e-01  2.67580122e-01  9.91457254e-02\n",
            " -7.83261433e-02  4.10892740e-02 -3.15037221e-01  2.05975637e-01\n",
            " -3.55794668e-01 -6.15333915e-01  9.95121822e-02 -2.48287782e-01\n",
            " -4.03497042e-03 -9.50843766e-02  4.70932797e-02 -5.56660354e-01\n",
            " -6.21537030e-01 -7.92744458e-01  3.01084965e-01 -3.24085981e-01\n",
            "  1.19331479e-01 -1.74839601e-01  4.94076572e-02  2.95995444e-01\n",
            " -4.59605932e-01 -1.95177242e-01 -2.53873259e-01  2.12419271e-01\n",
            "  1.84532061e-01 -5.33688307e-01 -1.05081156e-01  2.94583619e-01\n",
            " -8.33219171e-01 -9.63144779e-01 -9.57000911e-01 -8.45337868e-01\n",
            "  6.09982669e-01  5.05051970e-01 -1.28870860e-01  1.35004476e-01\n",
            " -2.94886976e-01  6.12230778e-01 -5.19383252e-01 -1.26083806e-01\n",
            " -5.30507267e-01 -5.77166677e-01 -6.21012628e-01  1.88742563e-01\n",
            "  6.63282096e-01  3.95969301e-01 -7.11457610e-01 -2.00106725e-01\n",
            "  4.85070914e-01  4.78976518e-01  3.01858573e-03  1.09103048e+00\n",
            " -2.08685901e-02  1.46742582e-01  1.39743760e-01 -4.45034564e-01\n",
            "  2.94522136e-01 -5.86887538e-01 -4.21239138e-01 -3.05410236e-01\n",
            " -1.09976792e+00 -8.00434873e-02  9.16814029e-01 -4.10235554e-01\n",
            " -1.37737429e+00 -4.59371716e-01 -1.46915480e-01 -6.73079848e-01\n",
            "  9.47711647e-01 -5.53089619e-01 -4.28550571e-01 -2.05269739e-01\n",
            "  2.53654629e-01  3.56386274e-01  4.62447137e-01  3.56955737e-01\n",
            " -1.13807440e-01  6.78821743e-01 -4.17779595e-01 -3.11268568e-01\n",
            " -2.98872977e-01  4.71550792e-01  6.48693383e-01 -3.66013825e-01\n",
            " -4.05527920e-01 -4.78279144e-01  2.44513184e-01  2.45983273e-01\n",
            "  1.10799469e-01  5.98745823e-01  1.17081106e-01 -1.51540041e-01\n",
            "  1.45668551e-01 -4.30412829e-01 -1.51079386e-01  1.26404893e+00\n",
            " -4.81337517e-01  1.44348168e+00 -3.55636925e-01  2.29145050e-01\n",
            " -1.04796901e-01  8.31979871e-01  7.72268236e-01 -5.31019986e-01\n",
            " -5.64369738e-01  5.02794147e-01 -1.49661198e-01 -1.27699792e+00\n",
            " -9.43129003e-01  4.10629809e-01  5.31995237e-01  6.81035891e-02\n",
            " -7.56149292e-02 -1.31650820e-01 -1.44370317e-01 -1.47911951e-01\n",
            "  2.54148722e-01 -1.17453766e+00 -4.83179279e-02 -3.55602264e-01\n",
            " -3.17777485e-01  5.67234540e-03  1.57523227e+00  8.88431132e-01\n",
            "  5.72320521e-01  2.68898427e-01 -5.76982684e-02  1.26121056e+00\n",
            " -1.44276753e-01 -2.53352165e-01  5.35118222e-01  7.51160800e-01\n",
            " -9.43924069e-01  8.68681312e-01  8.58703919e-04 -4.99720901e-01\n",
            " -6.73187077e-01 -2.13368282e-01  2.99257308e-01  4.13949251e-01\n",
            "  9.66982096e-02  1.07338488e+00  1.29762903e-01 -1.11246653e-01\n",
            " -6.20291233e-01  1.67305082e-01  1.46512939e-02 -4.26644862e-01\n",
            "  1.93598479e-01  2.17802078e-01 -1.53441399e-01 -7.33214021e-01\n",
            " -4.83393937e-01 -5.36713861e-02 -5.55786073e-01  1.31208569e-01\n",
            " -1.66739896e-01  7.23628700e-01 -1.22353531e-01 -2.95674711e-01\n",
            " -8.79835412e-02 -4.84484375e-01  1.65321216e-01 -4.86987978e-01\n",
            " -1.47212669e-01 -1.20190406e+00 -3.20209377e-02  8.82749036e-02\n",
            "  9.32432041e-02  4.26185802e-02 -4.39039201e-01 -9.81633127e-01\n",
            "  1.08841109e+00 -1.17028034e+00 -7.08803475e-01  4.39074934e-01\n",
            " -5.27613521e-01 -1.00992668e+00 -1.02849472e+00  4.75933030e-02\n",
            " -4.67369139e-01 -8.54000628e-01  1.43190116e-01  8.01121414e-01\n",
            " -6.15117550e-01 -1.41759574e+00  3.96752298e-01  5.86432874e-01\n",
            " -6.98901176e-01 -3.20745289e-01  1.03268409e+00 -1.70383346e+00\n",
            " -5.34656227e-01  5.16699731e-01 -1.10621536e+00 -4.75392640e-01\n",
            "  1.10720769e-01 -1.03432107e+00 -5.96579984e-02 -5.56520939e-01\n",
            "  4.62017059e-01  3.34096760e-01 -4.26750854e-02  6.58018649e-01\n",
            "  2.37163574e-01  3.59320343e-01  4.03034717e-01  1.54008293e+00\n",
            " -3.22026700e-01 -3.96218747e-01 -8.00728142e-01 -3.49993020e-01\n",
            " -7.68179476e-01 -8.33002627e-01 -6.19066179e-01 -5.12140989e-02\n",
            "  8.15447152e-01  6.36506617e-01  5.23187578e-01  2.98270494e-01\n",
            " -6.65145218e-01 -9.65569690e-02 -1.07671940e+00 -2.96323299e-01\n",
            " -1.59861460e-01  1.04468215e+00  1.07162702e+00  9.01240647e-01\n",
            " -2.85722762e-01  5.74721932e-01 -5.22451699e-01 -6.47164404e-01\n",
            "  1.24945574e-01 -2.64935076e-01 -5.41931450e-01 -2.27731928e-01\n",
            " -2.72647869e-02 -9.29802477e-01 -9.95185733e-01 -1.79678753e-01\n",
            "  6.29352480e-02  4.19814527e-01  1.15468407e+00  4.85228837e-01\n",
            "  3.99828583e-01  3.33518088e-01 -2.38265663e-01  2.91654259e-01\n",
            " -1.88552052e-01 -2.59676397e-01 -5.25504470e-01  4.31704700e-01\n",
            "  9.30795014e-01 -8.97127748e-01  7.04872668e-01 -3.20587784e-01\n",
            " -4.46830302e-01 -1.00823903e+00  2.56677233e-02  4.90723759e-01]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6qvIWuGomRec",
        "outputId": "57fa81e5-d92c-4df7-e6a5-c6e704ae83f6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "#Semantic Search \n",
        "\n",
        "#Import SciPy which is Scientific Programming Package, extension of Numpy, which provides working with multidimensional arrays\n",
        "import scipy\n",
        "\n",
        "#Enter Query to Search\n",
        "query = 'Election, Who won?'\n",
        "\n",
        "#Store as Variable\n",
        "queries = [query]\n",
        "\n",
        "#Encode Embedding of Query\n",
        "query_embeddings = model.encode(queries)\n",
        "\n",
        "#Get only three matches\n",
        "number_top_matches = 3\n",
        "print(\"Semantic Search Results\")\n",
        "\n",
        "#Calculate Distance using cosine of Sci Py Package\n",
        "for query, query_embedding in zip(queries, query_embeddings):\n",
        "  distances = scipy.spatial.distance.cdist([query_embedding], sentence_embeddings, \"cosine\")[0]\n",
        "  results = zip(range(len(distances)), distances)\n",
        "  results = sorted(results, key=lambda x:x[1])\n",
        "\n",
        "  print(\"\\n\\n=======\\n\\n\")\n",
        "  print(\"Query:\", query)\n",
        "  print(\"\\nTop 5 Simiarlity in Corupus\")\n",
        "\n",
        "  for idx, distance in results[0:number_top_matches]:\n",
        "    print(sentences[idx].strip(), \"(Cosine Score:%.4f)\"%(1-distance))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Semantic Search Results\n",
            "\n",
            "\n",
            "=======\n",
            "\n",
            "\n",
            "Query: Election, Who won?\n",
            "\n",
            "Top 5 Simiarlity in Corupus\n",
            "Who won the election? (Cosine Score:0.9581)\n",
            "How do we figure out who won the election? (Cosine Score:0.8196)\n",
            "Did everyone win in the election? (Cosine Score:0.7989)\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}