syntaxnet/examples/dragnn/interactive_text_analyzer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import ipywidgets as widgets\n",
    "import tensorflow as tf\n",
    "from IPython import display\n",
    "from dragnn.protos import spec_pb2\n",
    "from dragnn.python import graph_builder\n",
    "from dragnn.python import spec_builder\n",
    "from dragnn.python import load_dragnn_cc_impl  # This loads the actual op definitions\n",
    "from dragnn.python import render_parse_tree_graphviz\n",
    "from dragnn.python import visualization\n",
    "from google.protobuf import text_format\n",
    "from syntaxnet import load_parser_ops  # This loads the actual op definitions\n",
    "from syntaxnet import sentence_pb2\n",
    "from syntaxnet.ops import gen_parser_ops\n",
    "from tensorflow.python.platform import tf_logging as logging\n",
    "\n",
    "def load_model(base_dir, master_spec_name, checkpoint_name):\n",
    "    # Read the master spec\n",
    "    master_spec = spec_pb2.MasterSpec()\n",
    "    with open(os.path.join(base_dir, master_spec_name), \"r\") as f:\n",
    "        text_format.Merge(f.read(), master_spec)\n",
    "    spec_builder.complete_master_spec(master_spec, None, base_dir)\n",
    "    logging.set_verbosity(logging.WARN)  # Turn off TensorFlow spam.\n",
    "\n",
    "    # Initialize a graph\n",
    "    graph = tf.Graph()\n",
    "    with graph.as_default():\n",
    "        hyperparam_config = spec_pb2.GridPoint()\n",
    "        builder = graph_builder.MasterBuilder(master_spec, hyperparam_config)\n",
    "        # This is the component that will annotate test sentences.\n",
    "        annotator = builder.add_annotation(enable_tracing=True)\n",
    "        builder.add_saver()  # \"Savers\" can save and load models; here, we're only going to load.\n",
    "\n",
    "    sess = tf.Session(graph=graph)\n",
    "    with graph.as_default():\n",
    "        #sess.run(tf.global_variables_initializer())\n",
    "        #sess.run('save/restore_all', {'save/Const:0': os.path.join(base_dir, checkpoint_name)})\n",
    "        builder.saver.restore(sess, os.path.join(base_dir, checkpoint_name))\n",
    "        \n",
    "    def annotate_sentence(sentence):\n",
    "        with graph.as_default():\n",
    "            return sess.run([annotator['annotations'], annotator['traces']],\n",
    "                            feed_dict={annotator['input_batch']: [sentence]})\n",
    "    return annotate_sentence\n",
    "\n",
    "segmenter_model = load_model(\"data/en/segmenter\", \"spec.textproto\", \"checkpoint\")\n",
    "parser_model = load_model(\"data/en\", \"parser_spec.textproto\", \"checkpoint\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def annotate_text(text):\n",
    "    sentence = sentence_pb2.Sentence(\n",
    "        text=text,\n",
    "        token=[sentence_pb2.Token(word=text, start=-1, end=-1)]\n",
    "    )\n",
    "\n",
    "    # preprocess\n",
    "    with tf.Session(graph=tf.Graph()) as tmp_session:\n",
    "        char_input = gen_parser_ops.char_token_generator([sentence.SerializeToString()])\n",
    "        preprocessed = tmp_session.run(char_input)[0]\n",
    "    segmented, _ = segmenter_model(preprocessed)\n",
    "\n",
    "    annotations, traces = parser_model(segmented[0])\n",
    "    assert len(annotations) == 1\n",
    "    assert len(traces) == 1\n",
    "    return sentence_pb2.Sentence.FromString(annotations[0]), traces[0]\n",
    "annotate_text(\"John is eating pizza with a fork\"); None  # just make sure it works"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Interactive parse tree explorer\n",
    "Run the cell below, and then enter text in the interactive widget."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def _parse_tree_explorer():  # put stuff in a function to not pollute global scope\n",
    "    text = widgets.Text(\"John is eating pizza with anchovies\")\n",
    "    # Also try: John is eating pizza with a fork\n",
    "    display.display(text)\n",
    "    html = widgets.HTML()\n",
    "    display.display(html)\n",
    "\n",
    "    def handle_submit(sender):\n",
    "        del sender  # unused\n",
    "        parse_tree, trace = annotate_text(text.value)\n",
    "        html.value = u\"\"\"\n",
    "        <div style=\"max-width: 100%\">{}</div>\n",
    "        <style type=\"text/css\">svg {{ max-width: 100%; }}</style>\n",
    "        \"\"\".format(render_parse_tree_graphviz.parse_tree_graph(parse_tree))\n",
    "\n",
    "    text.on_submit(handle_submit)\n",
    "_parse_tree_explorer()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Interactive trace explorer\n",
    "Run the cell below, and then enter text in the interactive widget."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def _trace_explorer():  # put stuff in a function to not pollute global scope\n",
    "    text = widgets.Text(\"John is eating pizza with anchovies\")\n",
    "    display.display(text)\n",
    "\n",
    "    output = visualization.InteractiveVisualization()\n",
    "    display.display(display.HTML(output.initial_html()))\n",
    "\n",
    "    def handle_submit(sender):\n",
    "        del sender  # unused\n",
    "        parse_tree, trace = annotate_text(text.value)\n",
    "        display.display(display.HTML(output.show_trace(trace)))\n",
    "\n",
    "\n",
    "    text.on_submit(handle_submit)\n",
    "_trace_explorer()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}