n-gram-embedding.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load corpus"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(21573874, 8)"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import json\n",
    "import csv\n",
    "\n",
    "seasons = pd.read_csv(\"./data/bangumi.csv\", delimiter=\",\", encoding=\"utf-8\")\n",
    "episodes = pd.read_csv(\"./data/episode.csv\", delimiter=\",\", encoding=\"utf-8\")\n",
    "danmaku_complete = pd.read_csv(\"./data/danmaku_complete.csv\", delimiter=\"\\t\", encoding=\"utf-8\", quoting=csv.QUOTE_NONE, low_memory=False)\n",
    "danmaku_complete = danmaku_complete.fillna(-1)\n",
    "danmaku_complete.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "word segment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import jieba.posseg as segtool\n",
    "import re\n",
    "\n",
    "ACCEPTABLE_TYPE = {'n', 't', 's', 'f', 'v', 'a', 'b', 'z', 'e', 'y', 'o'}\n",
    "REPLACE_DICT = {\n",
    "    \"233+\": \"233\",\n",
    "    \"666+\": \"666\"\n",
    "}\n",
    "\n",
    "def check_type(word_type):\n",
    "    if word_type[0] in ACCEPTABLE_TYPE:\n",
    "        return True\n",
    "    else:\n",
    "        return False\n",
    "\n",
    "def check_replace(word):\n",
    "    for item in REPLACE_DICT.keys():\n",
    "        pattern = re.compile(item)\n",
    "        if re.match(pattern, word) is not None:\n",
    "            new_word = REPLACE_DICT[item]\n",
    "            return new_word\n",
    "    return word\n",
    "\n",
    "def word_segment(content):\n",
    "    words = []\n",
    "    results = segtool.cut(content)\n",
    "    for result in results:\n",
    "        result.word = check_replace(result.word)\n",
    "        if check_type(result.flag):\n",
    "            words.append(result.word)\n",
    "    return words"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "dataset definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.utils.data as data\n",
    "\n",
    "class DmDataset(data.Dataset):\n",
    "    def __init__(self, samples, vocab, context_size):\n",
    "        self.vocab = vocab\n",
    "        self.word_to_ix = {word: i for i, word in enumerate(vocab)}\n",
    "        self.samples = []\n",
    "        for i in range(len(samples)):\n",
    "            if len(samples[i]) < context_size:\n",
    "                seq = [self.word_to_ix[w] for w in samples[i]]\n",
    "                while len(seq) < context_size:\n",
    "                    seq.append(len(self.vocab))\n",
    "                current_sample = torch.LongTensor(seq)\n",
    "            else:\n",
    "                current_sample = torch.LongTensor([self.word_to_ix[w] for w in samples[i]])\n",
    "            self.samples.append(current_sample)\n",
    "        \n",
    "        \n",
    "    def word_to_ix(word):\n",
    "        if word in self.word_to_ix:\n",
    "            return self.word_to_ix[word]\n",
    "        else:\n",
    "            return len(self.vocab)\n",
    "        \n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        sample = self.samples[index]\n",
    "        return sample[:-1], sample[-1]\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.samples)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "data preprocess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Building prefix dict from the default dictionary ...\n",
      "Dumping model to file cache /var/folders/w9/1mf4l9j16f1_8x6smdnknjm00000gn/T/jieba.cache\n",
      "Loading model cost 1.330 seconds.\n",
      "Prefix dict has been built succesfully.\n"
     ]
    }
   ],
   "source": [
    "danmaku_selected = danmaku_complete[danmaku_complete['episode_id']==173248]\n",
    "\n",
    "vocab = set()\n",
    "context_size = 5\n",
    "samples = []\n",
    "\n",
    "for index, row in danmaku_selected.iterrows():\n",
    "    content = row['content']\n",
    "    words = word_segment(content)\n",
    "    for word in words:\n",
    "        vocab.add(word)\n",
    "    for i in range(0, len(words)-context_size):\n",
    "        samples.append(words[i:i+context_size])\n",
    "\n",
    "dm_set = DmDataset(samples, vocab, context_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Network definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch as t\n",
    "from torch import nn\n",
    "from torch.autograd import Variable as V\n",
    "import torch.nn.functional as F\n",
    "\n",
    "class NGramLanguageModeler(nn.Module):\n",
    "\n",
    "    def __init__(self, vocab_size, embedding_dim, context_size):\n",
    "        super(NGramLanguageModeler, self).__init__()\n",
    "        self.embeddings = nn.Embedding(vocab_size, embedding_dim)\n",
    "        self.linear1 = nn.Linear(context_size * embedding_dim, 128)\n",
    "        self.linear2 = nn.Linear(128, vocab_size)\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        embeds = self.embeddings(inputs).view(1,-1)\n",
    "        out = F.relu(self.linear1(embeds))\n",
    "        out = self.linear2(out)\n",
    "        log_probs = F.log_softmax(out, dim=1)\n",
    "        return log_probs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "training phase"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import torch\n",
    "import torch.utils.data as Data\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torch.autograd import Variable\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "%matplotlib inline\n",
    "\n",
    "is_ipython = 'inline' in matplotlib.get_backend()\n",
    "if is_ipython:\n",
    "    from IPython import display\n",
    "    \n",
    "def plot_durations(y):\n",
    "    plt.figure(1)\n",
    "    plt.clf()\n",
    "    plt.plot(y)\n",
    "\n",
    "#     plt.pause(0.001)  # pause a bit so that plots are updated\n",
    "    if is_ipython:\n",
    "        display.clear_output(wait=True)\n",
    "        display.display(plt.gcf())\n",
    "\n",
    "torch.manual_seed(1) \n",
    "\n",
    "EMBEDDING_DIM = 200\n",
    "batch_size=128\n",
    "losses = []\n",
    "loss_function = nn.NLLLoss()\n",
    "model = NGramLanguageModeler(len(vocab), EMBEDDING_DIM, context_size-1)\n",
    "print(model)\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.001)\n",
    "dm_dataloader = Data.DataLoader(\n",
    "    dataset=dm_set,\n",
    "    batch_size=128,\n",
    "    shuffle=True,\n",
    "    drop_last=True,\n",
    "    num_workers=8\n",
    ")\n",
    "\n",
    "if torch.cuda.is_available():\n",
    "    model.cuda()\n",
    "    \n",
    "loss_series = []\n",
    "plt.ion()\n",
    "\n",
    "for epoch in range(10):\n",
    "    for batch_idx, (context_feature, label) in enumerate(dm_dataloader):\n",
    "        total_loss = 0\n",
    "        for step in range(batch_size):    \n",
    "            context_feature_ = Variable(torch.LongTensor(context_feature[step]))\n",
    "            label_ = Variable(torch.LongTensor([label[step]]))\n",
    "            \n",
    "            if torch.cuda.is_available():\n",
    "                context_feature_ = context_feature_.cuda()\n",
    "                label_ = label_.cuda()\n",
    "        \n",
    "            model.zero_grad()\n",
    "            log_probs = model(context_feature_)\n",
    "            loss = loss_function(log_probs, label_)\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            total_loss += loss.cpu().data.numpy()\n",
    "        \n",
    "        loss_series.append(total_loss/batch_size)\n",
    "        plot_durations(loss_series)\n",
    "#         print(\"epoch: %d loss: %.4f\" % (epoch, total_loss))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([1, 200])\n"
     ]
    }
   ],
   "source": [
    "context_feature_ = Variable(torch.LongTensor([1]))\n",
    "em = model.embeddings(context_feature_)\n",
    "print(em.data.size())"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:research]",
   "language": "python",
   "name": "conda-env-research-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}