tutorials/00_basic_usage.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# clone and pip install the package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "import kenlm\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(371, 29)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[-22, -19, -19, ..., -22, -19,   0],\n",
       "       [-23, -19, -19, ..., -22, -19,   0],\n",
       "       [-25, -21, -20, ..., -23, -20,   0],\n",
       "       ...,\n",
       "       [-16, -17, -19, ..., -20, -15,   0],\n",
       "       [-15, -16, -18, ..., -19, -14,   0],\n",
       "       [-16, -16, -18, ..., -19, -15,   0]])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TEST_LOGIT_PATH = \"../pyctcdecode/tests/sample_data/libri_logits.json\"\n",
    "with open(TEST_LOGIT_PATH) as f:\n",
    "    logits = np.array(json.load(f))\n",
    "    \n",
    "print(logits.shape)\n",
    "logits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = [\n",
    "    \" \", \n",
    "    \"a\", \"b\", \"c\", \"d\", \"e\", \"f\", \"g\", \"h\", \"i\", \"j\", \"k\", \"l\", \"m\", \n",
    "    \"n\", \"o\", \"p\", \"q\", \"r\", \"s\", \"t\", \"u\", \"v\", \"w\", \"x\", \"y\", \"z\", \n",
    "    \"'\",\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def _greedy_decode(logits):\n",
    "    \"\"\"Decode argmax of logits and squash in CTC fashion.\"\"\"\n",
    "    label_dict = {n: c for n, c in enumerate(labels)}\n",
    "    prev_c = None\n",
    "    out = []\n",
    "    for n in logits.argmax(axis=1):\n",
    "        c = label_dict.get(n, \"\")  # if not in labels, then assume it's ctc blank char\n",
    "        if c != prev_c:\n",
    "            out.append(c)\n",
    "        prev_c = c\n",
    "    return \"\".join(out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'i have a good deal of will you remember and what i have set my mind upon no doubt i shall some day achieve'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "_greedy_decode(logits)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'i have a good deal of will you remember and what i have set my mind upon no doubt i shall some day achieve'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from pyctcdecode import build_ctcdecoder\n",
    "\n",
    "decoder = build_ctcdecoder(labels)\n",
    "decoder.decode(logits)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}