Upload qc_viewer_colab.ipynb with huggingface_hub

qc_viewer_colab.ipynb

@@ -0,0 +1,226 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# VoxPopuli QC Samples V3 - CER Viewer\n",
+    "\n",
+    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/toth235a/voxpopuli-qc-samples-v3/blob/main/qc_viewer_colab.ipynb)\n",
+    "\n",
+    "Quality control samples categorized by Character Error Rate (CER) between Whisper and Parakeet transcriptions.\n",
+    "\n",
+    "**CER Bins:**\n",
+    "- `cer_02_05`: 2-5% CER (good reference)\n",
+    "- `cer_05_10`: 5-10% CER (slight difference)\n",
+    "- `cer_10_20`: 10-20% CER (moderate difference)\n",
+    "- `cer_20_30`: 20-30% CER (significant difference)\n",
+    "- `cer_30_50`: 30-50% CER (very different)\n",
+    "- `cer_50_plus`: 50%+ CER (extremely different)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Install dependencies\n",
+    "!pip install -q datasets ipywidgets"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from datasets import load_dataset\n",
+    "from IPython.display import display, Audio, HTML, clear_output\n",
+    "import ipywidgets as widgets\n",
+    "\n",
+    "# Load dataset from HuggingFace\n",
+    "ds = load_dataset(\"toth235a/voxpopuli-qc-samples-v3\", split=\"train\")\n",
+    "print(f\"Loaded {len(ds)} samples\")\n",
+    "print(f\"Columns: {ds.column_names}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Group by CER bin\n",
+    "bins = {}\n",
+    "for i, sample in enumerate(ds):\n",
+    "    bin_name = sample['cer_bin']\n",
+    "    if bin_name not in bins:\n",
+    "        bins[bin_name] = []\n",
+    "    bins[bin_name].append((i, sample))\n",
+    "\n",
+    "bin_order = ['cer_02_05', 'cer_05_10', 'cer_10_20', 'cer_20_30', 'cer_30_50', 'cer_50_plus']\n",
+    "print(f\"Bins: {[b for b in bin_order if b in bins]}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def display_sample(idx, sample):\n",
+    "    \"\"\"Display a single sample with full transcripts (NO truncation).\"\"\"\n",
+    "    \n",
+    "    # Header with metadata\n",
+    "    html = f\"\"\"\n",
+    "    <div style=\"border: 2px solid #333; padding: 15px; margin: 10px 0; border-radius: 8px; background: #f9f9f9;\">\n",
+    "        <h3 style=\"margin-top: 0;\">Sample: {sample['id']}</h3>\n",
+    "        <table style=\"width: 100%; margin-bottom: 15px;\">\n",
+    "            <tr>\n",
+    "                <td><b>CER Bin:</b> {sample['cer_bin']}</td>\n",
+    "                <td><b>CER:</b> {sample['cer']:.1%}</td>\n",
+    "                <td><b>Duration:</b> {sample['duration']:.1f}s</td>\n",
+    "            </tr>\n",
+    "        </table>\n",
+    "    </div>\n",
+    "    \"\"\"\n",
+    "    display(HTML(html))\n",
+    "    \n",
+    "    # Audio player\n",
+    "    audio_data = sample['audio']\n",
+    "    display(Audio(audio_data['array'], rate=audio_data['sampling_rate']))\n",
+    "    \n",
+    "    # Full transcripts - NO TRUNCATION\n",
+    "    whisper = sample['whisper_text']\n",
+    "    parakeet = sample['parakeet_text']\n",
+    "    \n",
+    "    transcript_html = f\"\"\"\n",
+    "    <div style=\"margin-top: 15px;\">\n",
+    "        <div style=\"background: #e8f4e8; padding: 12px; border-radius: 6px; margin-bottom: 10px;\">\n",
+    "            <b style=\"color: #2d5a2d;\">Whisper ({len(whisper)} chars):</b>\n",
+    "            <p style=\"white-space: pre-wrap; word-wrap: break-word; margin: 8px 0 0 0; font-family: monospace; font-size: 13px;\">{whisper}</p>\n",
+    "        </div>\n",
+    "        <div style=\"background: #e8e8f4; padding: 12px; border-radius: 6px;\">\n",
+    "            <b style=\"color: #2d2d5a;\">Parakeet ({len(parakeet)} chars):</b>\n",
+    "            <p style=\"white-space: pre-wrap; word-wrap: break-word; margin: 8px 0 0 0; font-family: monospace; font-size: 13px;\">{parakeet}</p>\n",
+    "        </div>\n",
+    "    </div>\n",
+    "    \"\"\"\n",
+    "    display(HTML(transcript_html))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Interactive viewer\n",
+    "bin_dropdown = widgets.Dropdown(\n",
+    "    options=[(f\"{b} ({len(bins.get(b, []))} samples)\", b) for b in bin_order if b in bins],\n",
+    "    description='CER Bin:',\n",
+    "    style={'description_width': '80px'}\n",
+    ")\n",
+    "\n",
+    "sample_slider = widgets.IntSlider(\n",
+    "    min=0, max=9, value=0,\n",
+    "    description='Sample #:',\n",
+    "    style={'description_width': '80px'}\n",
+    ")\n",
+    "\n",
+    "output = widgets.Output()\n",
+    "\n",
+    "def update_display(*args):\n",
+    "    with output:\n",
+    "        clear_output(wait=True)\n",
+    "        bin_name = bin_dropdown.value\n",
+    "        idx = sample_slider.value\n",
+    "        if bin_name in bins and idx < len(bins[bin_name]):\n",
+    "            ds_idx, sample = bins[bin_name][idx]\n",
+    "            display_sample(ds_idx, sample)\n",
+    "\n",
+    "def on_bin_change(*args):\n",
+    "    bin_name = bin_dropdown.value\n",
+    "    sample_slider.max = len(bins.get(bin_name, [])) - 1\n",
+    "    sample_slider.value = 0\n",
+    "    update_display()\n",
+    "\n",
+    "bin_dropdown.observe(on_bin_change, names='value')\n",
+    "sample_slider.observe(update_display, names='value')\n",
+    "\n",
+    "# Navigation buttons\n",
+    "prev_btn = widgets.Button(description='< Prev', button_style='info')\n",
+    "next_btn = widgets.Button(description='Next >', button_style='info')\n",
+    "\n",
+    "def on_prev(b):\n",
+    "    if sample_slider.value > 0:\n",
+    "        sample_slider.value -= 1\n",
+    "\n",
+    "def on_next(b):\n",
+    "    if sample_slider.value < sample_slider.max:\n",
+    "        sample_slider.value += 1\n",
+    "\n",
+    "prev_btn.on_click(on_prev)\n",
+    "next_btn.on_click(on_next)\n",
+    "\n",
+    "# Layout\n",
+    "controls = widgets.HBox([bin_dropdown, sample_slider, prev_btn, next_btn])\n",
+    "display(controls)\n",
+    "display(output)\n",
+    "\n",
+    "# Initial display\n",
+    "on_bin_change()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "---\n",
+    "## Browse All Samples in a Bin"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Change this to view different bins\n",
+    "BIN_TO_SHOW = 'cer_50_plus'\n",
+    "\n",
+    "for ds_idx, sample in bins.get(BIN_TO_SHOW, []):\n",
+    "    display_sample(ds_idx, sample)\n",
+    "    display(HTML(\"<hr style='border: 1px dashed #ccc; margin: 20px 0;'>\"))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Summary table\n",
+    "print(\"All samples summary:\")\n",
+    "print(f\"{'ID':<20} {'Bin':<15} {'CER':>8} {'Duration':>10}\")\n",
+    "print(\"-\" * 55)\n",
+    "for sample in sorted(ds, key=lambda x: (x['cer_bin'], x['cer'])):\n",
+    "    print(f\"{sample['id']:<20} {sample['cer_bin']:<15} {sample['cer']:>7.1%} {sample['duration']:>9.1f}s\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "name": "python",
+   "version": "3.10.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}