Upload folder using huggingface_hub

README.md

@@ -1,3 +1,219 @@
+# Semantic Interruption Detection Benchmark
+
+A multilingual benchmark for evaluating **semantic interruption detection** in conversational speech. The dataset contains annotated audio recordings in English and Mandarin Chinese, along with a silence/noise negative set, designed to test models that decide *when* (and *whether*) to interrupt a speaker.
+
 ---
-license: cc-by-4.0
+
+## Overview
+
+In spoken dialogue systems and voice assistants, knowing when to interrupt a speaker — and detecting interruption signals in real time — is a critical yet under-studied task. This benchmark provides:
+
+- **3,700 audio clips** with frame-level interruption timestamps
+- **Bilingual coverage**: English and Mandarin Chinese conversational speech
+- **Negative samples**: 500 silence and environmental noise clips for robustness evaluation
+- **Three standardized metrics**: FIR, IRL, and APT (defined below)
+
 ---
+
+## Dataset Statistics
+
+| Split | Language | Total Samples | With Interruption | Without Interruption |
+|-------|----------|:-------------:|:-----------------:|:--------------------:|
+| `en_test_lines.jsonl` | English | 1,600 | 1,100 (68.8%) | 500 (31.3%) |
+| `zh_test_lines.jsonl` | Chinese | 1,600 | 1,100 (68.8%) | 500 (31.3%) |
+| `silence_noise_test.jsonl` | — | 500 | 0 (0%) | 500 (100%) |
+| **Total** | | **3,700** | **2,200** | **1,500** |
+
+**Audio duration:**
+- English: 0.36s – 22.0s (avg 3.74s)
+- Chinese: similar range (avg 3.85s)
+- Silence/noise: 0.09s – 182.0s (avg 13.0s)
+
+---
+
+## Directory Structure
+
+```
+Semantic_Interaption/
+├── README.md
+├── test.ipynb                        # Evaluation notebook
+└── test_wavs/
+    ├── en_test_lines.jsonl           # English ground-truth annotations
+    ├── zh_test_lines.jsonl           # Chinese ground-truth annotations
+    ├── silence_noise_test.jsonl      # Silence / noise annotations
+    ├── en/                           # English WAV files (1,600)
+    ├── zh/                           # Chinese WAV files (1,600)
+    └── silence_or_noise/             # Silence / noise WAV files (500)
+```
+
+---
+
+## Annotation Format
+
+Each `.jsonl` file contains one JSON object per line.
+
+### Speech splits (`en_test_lines.jsonl`, `zh_test_lines.jsonl`)
+
+```json
+{
+  "audio":           "MDT_ASR_AI106_A0016_S0005_0_G3265_0318890-0324670.wav",
+  "total_nonbreak":  false,
+  "duration":        5.78,
+  "break_time":      0.125,
+  "text_with_break": "<break> Experiencing different cultures, learning new languages, ..."
+}
+```
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `audio` | `str` | Filename of the corresponding WAV file |
+| `total_nonbreak` | `bool` | `true` if no interruption occurs in the entire clip |
+| `duration` | `float` | Total audio duration in seconds |
+| `break_time` | `float` | Timestamp (seconds) of the first interruption signal; `-1` when `total_nonbreak=true` |
+| `text_with_break` | `str` | Transcript with `<break>` token inserted at the interruption point |
+
+### Silence / noise split (`silence_noise_test.jsonl`)
+
+All samples have `total_nonbreak=true` and `text_with_break=null`. These clips contain no speech and serve as a robustness / false-alarm test.
+
+```json
+{
+  "audio":           "wind_wind_heavy_clean_stdy_air.wav",
+  "total_nonbreak":  true,
+  "duration":        152.52,
+  "break_time":      152.52,
+  "text_with_break": null
+}
+```
+
+---
+
+## Evaluation Metrics
+
+The benchmark defines three complementary metrics. Reference implementation is in `test.ipynb`.
+
+### FIR — False Interruption Rate
+
+The fraction of non-interruption samples for which the model incorrectly predicts an interruption.
+
+```
+FIR = |false positives| / |total_nonbreak samples|
+```
+
+Lower is better.
+
+### IRL — Interruption Response Latency
+
+Average absolute timing error (in seconds) over true-positive detections (model correctly identifies a break and the timestamp is within 50 ms of ground truth).
+
+```
+IRL = mean(|predicted_break_time - gt_break_time|)   [over true positives]
+```
+
+Lower is better.
+
+### APT — Average Penalty Time
+
+A unified penalty that captures both detection errors and timing inaccuracies:
+
+| Prediction vs. Ground Truth | Penalty |
+|-----------------------------|---------|
+| Both predict no break | 0 |
+| Predicted time within ±50 ms of ground truth | 0 |
+| False alarm (predict break, no GT break) | Full clip duration |
+| Missed detection (no prediction, GT has break) | `duration − break_time` |
+| Late detection | `predicted_time − break_time` |
+
+```
+APT = mean(penalty)   [over all samples]
+```
+
+Lower is better.
+
+---
+
+## Download
+
+The full dataset (audio + annotations) is hosted on Hugging Face:
+
+**[https://huggingface.co/datasets/kxxia/SID-bench](https://huggingface.co/datasets/kxxia/SID-bench)**
+
+```bash
+# Option 1: huggingface_hub CLI
+huggingface-cli download kxxia/SID-bench --repo-type dataset --local-dir ./SID-bench
+
+# Option 2: Python
+from huggingface_hub import snapshot_download
+snapshot_download(repo_id="kxxia/SID-bench", repo_type="dataset", local_dir="./SID-bench")
+```
+
+---
+
+## Quick Start
+
+```python
+import json
+
+# Load ground-truth annotations
+with open("test_wavs/en_test_lines.jsonl") as f:
+    en_samples = [json.loads(line) for line in f]
+
+# Each sample
+sample = en_samples[0]
+print(sample["audio"])           # WAV filename
+print(sample["total_nonbreak"])  # bool
+print(sample["break_time"])      # float (seconds)
+print(sample["text_with_break"]) # transcript with <break> marker
+```
+
+Run `test.ipynb` to reproduce the evaluation pipeline and compute FIR / IRL / APT against your own model predictions.
+
+### Prediction format
+
+Your model should produce a JSONL file with one prediction per line:
+
+```json
+{"audio": "MDT_ASR_AI106_A0016_S0005_0_G3265_0318890-0324670.wav", "total_nonbreak": false, "break_time": 0.13}
+```
+
+---
+
+## Use Cases
+
+- **Conversational AI / dialogue systems**: training models to detect natural turn-taking signals
+- **Voice assistants**: knowing when a user intends to interrupt or be interrupted
+- **Speech-to-speech translation**: preserving interruption timing across languages
+- **Contact center analytics**: segmenting overlapping speech
+
+---
+
+## License
+
+Please check the `LICENSE` file for usage terms. The audio data may originate from multiple source corpora; refer to their respective licenses before use in commercial applications.
+
+---
+
+## Acknowledgement
+
+The English conversational speech data used in this benchmark were derived from the **English Duplex Conversation Training Dataset** provided by MagicData.
+
+> MagicData. (2025). *MDT-AF069 English Duplex Conversation Training Dataset*.
+>https://www.magicdatatech.com/datasets/asr/mdt-af069-multi-stream-spontaneous-conversation-training-datasets-english-1733294791
+
+---
+
+## Citation
+
+If you use this benchmark in your research, please cite:
+
+```bibtex
+@misc{semantic_interruption_benchmark,
+      title={Semantic-Aware Interruption Detection in Spoken Dialogue Systems: Benchmark, Metric, and Model}, 
+      author={Kangxiang Xia and Bingshen Mu and Xian Shi and Jin Xu and Lei Xie},
+      year={2026},
+      eprint={2603.24144},
+      archivePrefix={arXiv},
+      primaryClass={cs.SD},
+      url={https://arxiv.org/abs/2603.24144}, 
+}
+```

test.ipynb

@@ -0,0 +1,234 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from tqdm import tqdm\n",
+    "from json_repair import repair_json\n",
+    "import os\n",
+    "import json\n",
+    "import random\n",
+    "import re\n",
+    "\n",
+    "def readListFromFile(path: str) -> list:\n",
+    "    \"\"\"从文件中逐行读取内容到列表\"\"\"\n",
+    "    lines = []\n",
+    "    try:\n",
+    "        with open(path, 'r', encoding='utf-8') as f:\n",
+    "            lines = [line.strip() for line in f]\n",
+    "    except UnicodeDecodeError:\n",
+    "        print(f\"{path} utf-8 打开失败，尝试使用gbk\")\n",
+    "        try:\n",
+    "            with open(path, 'r', encoding='gbk') as f:\n",
+    "                lines = [line.strip() for line in f]\n",
+    "        except Exception as e:\n",
+    "            print(f\"读取文件失败: {e}\")\n",
+    "    except FileNotFoundError:\n",
+    "        print(f\"错误: 文件未找到 - {path}\")\n",
+    "    \n",
+    "    if path.endswith('.jsonl'):\n",
+    "        try:\n",
+    "            json_lines = [json.loads(line) for line in lines]\n",
+    "            return json_lines\n",
+    "        except Exception as e:\n",
+    "            print(f\"JSON 解析失败: {e}\")\n",
+    "            \n",
+    "    return lines\n",
+    "\n",
+    "def saveLstAsFile(data_list: list, path: str, overwrite: bool = False):\n",
+    "    \"\"\"将列表内容逐行写入文件\"\"\"\n",
+    "    if len(data_list) == 0:\n",
+    "        print(\"列表为空，无法写入文件\")\n",
+    "        return\n",
+    "    \n",
+    "    if isinstance(data_list[0],dict):\n",
+    "        data_list = [json.dumps(item, ensure_ascii=False) for item in data_list]\n",
+    "\n",
+    "    if os.path.exists(path) and not overwrite:\n",
+    "        new_path = f'{path}_{int(time.time())}.bak'\n",
+    "        print(f'{path} 已存在，内容将写入到 {new_path}')\n",
+    "        path = new_path\n",
+    "    with open(path, 'w', encoding='utf-8') as f:\n",
+    "        for item in data_list:\n",
+    "            f.write(str(item).strip() + '\\n')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "zh_gt_labs_lines = readListFromFile('./test_wavs/zh_test_lines.jsonl')\n",
+    "en_gt_labs_lines = readListFromFile('./test_wavs/en_test_lines.jsonl')\n",
+    "ns_gt_labs_lines = readListFromFile('./test_wavs/silence_noise_test.jsonl')\n",
+    "\n",
+    "audio2lines = {\n",
+    "    line['audio']: line\n",
+    "    for line in zh_gt_labs_lines + en_gt_labs_lines + ns_gt_labs_lines\n",
+    "}"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "`./zh_our_res.jsonl` 文件中，每一行都是一个 JSON 对象，例如：\n",
+    "\n",
+    "```jsonl\n",
+    "{\"audio\": \"00885_00886_00885_701.wav\", \"total_nonbreak\": true, \"break_time\": -1}\n",
+    "{\"audio\": \"00883_00884_00883_172.wav\", \"total_nonbreak\": false, \"break_time\": 0.64}\n",
+    "```\n",
+    "\n",
+    "### 字段说明\n",
+    "\n",
+    "```json\n",
+    "{\n",
+    "  \"audio\": \"音频文件名\",\n",
+    "  \"total_nonbreak\": \"是否整段音频都未触发打断信号\",\n",
+    "  \"break_time\": \"触发打断信号的时间点（单位：秒）\"\n",
+    "}\n",
+    "```\n",
+    "\n",
+    "### 规则说明\n",
+    "\n",
+    "- `audio`：音频文件名。\n",
+    "- `total_nonbreak`：\n",
+    "  - `true` 表示整段音频没有触发打断信号；\n",
+    "  - `false` 表示音频中触发了打断信号。\n",
+    "- `break_time`：\n",
+    "  - 当 `total_nonbreak = true` 时，`break_time` 通常为 `-1`；\n",
+    "  - 当 `total_nonbreak = false` 时，`break_time` 表示首次触发打断信号的时间点，单位为秒。\n",
+    "\n",
+    "---\n",
+    "\n",
+    "Each line in `./zh_our_res.jsonl` is a JSON object, for example:\n",
+    "\n",
+    "```jsonl\n",
+    "{\"audio\": \"00885_00886_00885_701.wav\", \"total_nonbreak\": true, \"break_time\": -1}\n",
+    "{\"audio\": \"00883_00884_00883_172.wav\", \"total_nonbreak\": false, \"break_time\": 0.64}\n",
+    "```\n",
+    "\n",
+    "### Field Description\n",
+    "\n",
+    "```json\n",
+    "{\n",
+    "  \"audio\": \"Audio filename\",\n",
+    "  \"total_nonbreak\": \"Whether the entire audio contains no interruption signal\",\n",
+    "  \"break_time\": \"Timestamp of the interruption signal (in seconds)\"\n",
+    "}\n",
+    "```\n",
+    "\n",
+    "### Rules\n",
+    "\n",
+    "- `audio`: the audio filename.\n",
+    "- `total_nonbreak`:\n",
+    "  - `true` means no interruption signal was triggered throughout the entire audio;\n",
+    "  - `false` means an interruption signal was triggered in the audio.\n",
+    "- `break_time`:\n",
+    "  - When `total_nonbreak = true`, `break_time` is usually `-1`;\n",
+    "  - When `total_nonbreak = false`, `break_time` indicates the timestamp of the first interruption signal, in seconds.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "test_result_file = './zh_our_res.jsonl'\n",
+    "test_result_lines = readListFromFile(test_result_file)\n",
+    "\n",
+    "false_alarm_num = 0\n",
+    "for line in test_result_lines:\n",
+    "    audio = line['audio']\n",
+    "    labs = audio2lines[audio]\n",
+    "\n",
+    "    test_break_time = line['break_time']\n",
+    "    test_nonbreak = line['total_nonbreak']\n",
+    "\n",
+    "    gt_break_time = labs['break_time']\n",
+    "    gt_nonbreak = labs['total_nonbreak']\n",
+    "    duration = labs['duration']\n",
+    "\n",
+    "    line['duration'] = duration\n",
+    "    line['gt_break_time'] = gt_break_time\n",
+    "    line['gt_nonbreak'] = gt_nonbreak\n",
+    "\n",
+    "    if test_break_time <= gt_break_time + 0.05 and test_break_time >= gt_break_time - 0.05: #结果打断时间和真实打断时间相近\n",
+    "        line['advance_time'] = 0\n",
+    "        line['delay_time'] = 0\n",
+    "        line['penalty_time'] = 0\n",
+    "        continue\n",
+    "    elif gt_nonbreak and test_nonbreak:   #真实不打断，结果不打断\n",
+    "        line['advance_time'] = 0\n",
+    "        line['delay_time'] = 0\n",
+    "        line['penalty_time'] = 0\n",
+    "        continue\n",
+    "    elif gt_nonbreak and not test_nonbreak: #真实不打断，结果打断\n",
+    "        false_alarm_num += 1\n",
+    "        line['advance_time'] = 0\n",
+    "        line['delay_time'] = 0\n",
+    "        line['penalty_time'] = duration\n",
+    "    elif not gt_nonbreak and test_nonbreak: #真实打断，结果不打断\n",
+    "        line['advance_time'] = 0\n",
+    "        line['delay_time'] = 0\n",
+    "        line['penalty_time'] = (duration-gt_break_time)\n",
+    "    elif not gt_nonbreak and not test_nonbreak: #真实打断，结果打断\n",
+    "        if test_break_time >= gt_break_time:\n",
+    "            line['advance_time'] = 0\n",
+    "            line['delay_time'] = (test_break_time - gt_break_time)\n",
+    "            line['penalty_time'] = (test_break_time - gt_break_time)\n",
+    "        elif test_break_time < gt_break_time:\n",
+    "            false_alarm_num += 1\n",
+    "            line['advance_time'] = 0\n",
+    "            line['delay_time'] = 0\n",
+    "            line['penalty_time'] = duration\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "FIR: 0.1375 IRL: 0.33837436932391457 APT: 0.6561522135416662\n"
+     ]
+    }
+   ],
+   "source": [
+    "delay_items = [line['delay_time'] for line in test_result_lines if line['delay_time'] > 0]\n",
+    "penalty_time_items = [line['penalty_time'] for line in test_result_lines]\n",
+    "\n",
+    "print(\"FIR:\",false_alarm_num/len(test_result_lines), \"IRL:\", sum(delay_items)/len(delay_items), \"APT:\", sum(penalty_time_items)/len(penalty_time_items))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "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": "3.9.23"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

test_wavs/en/00879_00880_00879_610.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3af8c4079776fb3cfe1ed76b1caec4146b6c27adf85b1757ab34b345eaf24056
+size 2111372

test_wavs/en/00879_00880_00879_677.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a19118d05dd5f5db97b13fa360ebd9a281b2d1d2db9e1da6633ae95753873e3b
+size 499436

test_wavs/en/00879_00880_00880_263.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a1c15d6ac73a115c04cba19fc8286c907955621e007f7600ecb3d28894847b4
+size 231884

test_wavs/en/00883_00884_00883_180.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec745a545135fdb428f94dcc63680f0c37609c1afc5e3f2e94de193cfd1ea346
+size 1283372

test_wavs/en/00883_00884_00883_190.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78d72c38283e1be5a08a2deb3e31e99c509aeeda58e50b35d17bc33d4733d131
+size 935660

test_wavs/en/00883_00884_00883_200.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4730f650f14f17c400653d7bc8357ff6523be666f1633f360908a0e13d608e2f
+size 1047308

test_wavs/en/00883_00884_00883_403.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89f72334c21853348c09a3ff647ffb4db85e0b0f84362cf6e5fc2d9e64ea10b4
+size 1370924

test_wavs/en/00883_00884_00883_614.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4727679698d0e675ab695d980fb3503a7cc9111d64701a892c750a950802cbf5
+size 150476

test_wavs/en/00883_00884_00884_161.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4d7b6c0115cc70396e464ec72c9ab83c9bfb83a685cedf8bd3d52b903e21dba
+size 844172

test_wavs/en/00883_00884_00884_299.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0d5d9c5913bbd96e857ae4881f5068aebf3023d996bdeb5549f3538dbb04c73
+size 689230

test_wavs/en/00883_00884_00884_308.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83791a2c0c6bed3d3a82b88d9ec55b8ab739de8749b0e290663e8110b56b10c1
+size 953228

test_wavs/en/00883_00884_00884_422.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a2c5c1d1419c34631097d231850a665ab5dcfc82a97ee36ce3dfc3f240246cd
+size 611468

test_wavs/en/00883_00884_00884_428.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7658b486271c8069f8b0a67896f66619143283501c1aa4597ba75187b8f47d56
+size 124652

test_wavs/en/00883_00884_00884_622.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a996163883ffc949f24851e16cbba2ddb0d3a378335f71208d31bf6fe1e673b6
+size 399500

test_wavs/en/00883_00884_00884_673.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:959b04df7d438a7c4008946f34e470936732a34683f76592408c5ad22ab22ac3
+size 384620

test_wavs/en/00883_00884_00884_719.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48c6d0e529071c8aab7c90f7a27575ecf0af821a14e125054bf11ab5a10da769
+size 234092

test_wavs/en/00885_00886_00885_240.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a631910c276f96dfceb987605f08d16897fb3d9098e5c2356bcc653e54b229e
+size 708138

test_wavs/en/00885_00886_00885_262.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:420bcc7258db34753790d65e209822031be0f8b35a2e54345866f71a5998e8c9
+size 269516

test_wavs/en/00885_00886_00885_276.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8ddf19f81aefd5ad945c3f8b69ca1e3335a4512bd5fced662642519e4a6b661
+size 293612

test_wavs/en/00885_00886_00885_292.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e28687608951afd6a301ec5471fce031a5663aed89020fd550cfc97b2180826
+size 654188

test_wavs/en/00885_00886_00885_362.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:225d45cc6ca8f8bbaf58b7d8a0f422be9195171c45a12f251286d9dab03dd18d
+size 419276

test_wavs/en/00885_00886_00885_392.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45b13e4d5c3f07d1218c518b6fce6005a5c8f5d979e75ca2b04a84599435b8db
+size 187820

test_wavs/en/00885_00886_00885_441.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0a52772891c44ef9b2fa6a72af601d6fb13a39a23785beed466aefe23d293e6
+size 580556

test_wavs/en/00885_00886_00885_540.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bfa9d1f7faf983f0be9114902f666fde6c26454e852543f936dfd7fd549ee76
+size 251468

test_wavs/en/00885_00886_00885_597.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:872ec828a4b0084407050939eb51b578991f51759e226e93b290cb7421f412e2
+size 561836

test_wavs/en/CTS-CN-F2F-2018-11-19_248_clipped_61.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da08d4c882c9d024b9964d4135e5ab9e4d76092b58b6edd748c9de853de6b37b
+size 48140

test_wavs/en/CTS-CN-F2F-2018-11-19_251_clipped_9.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f471a743528c0e2c1f4b5eb9d5a50371a4dca572c733b893cddee1e5961e64dc
+size 31692

test_wavs/en/CTS-CN-F2F-2018-11-19_256_clipped_266.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76846973371610b649708606a50078b9e737e831ddf2fb847e8a83d467f75e15
+size 342924

test_wavs/en/CTS-CN-F2F-2018-11-19_256_clipped_77.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61ed1ff38e5763ec12f8fe8136f590497189e67318779148879134ef6ef9cf1f
+size 40076

test_wavs/en/CTS-CN-F2F-2018-11-19_259_clipped_1.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f29746da0d9551b5ba784282e36e72c908db6618d89010251d18cb2fea439fb4
+size 239596

test_wavs/en/CTS-CN-F2F-2018-11-19_259_clipped_97.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6a9a52ed9d0f31ebda8c3aee46a59a48e058a1f184ad3f977b6fd005d6d066d
+size 458668

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0164_0714376-0716218.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3f7c9c5ad2f2e94d1710bbfa10021b74f42ca5a494e3cbe59619937caadc069
+size 57422

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0164_0869873-0870746.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:331934c01b85a931f790570a3e8d5c5d5e3d1e9ef0340d931ffb8c6b42c8035a
+size 28750

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0027830-0035380.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e97f9d86180fe1ecbd2cdca9fcdf4e13b53a06770077b085fd39d1b6c53682aa
+size 241742

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0403085-0405688.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4613db733a6d5f1c502e16e797cefb121824c5411a21239ab50b389cbfde03f2
+size 81998

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0511073-0513812.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdbd3269ba3f661a979a2d2b3fddca96af4a1347512dd0651801373659887154
+size 90190

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0549120-0550672.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd53e31dd5671f100840d9e4400a3dca333c780b6807c98ced22afd71251ce02
+size 53326

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0672072-0672518.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c93caccb5bed7a0ebc1ecec615f95fac8ce094bc770116a4bf574da10b8866cd
+size 16462

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0761730-0768140.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22c2bbc07a33c0e8aef5dca0bb346e85094f47c26ddb079ec9db33a3d5de7b3c
+size 204878

test_wavs/en/MDT_ASR_AF069A_A0006_S011_0_G0165_0795750-0798710.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ad451109dd1a9f67966e49129fc3c19f33198216d28c40b5d052e453f25b048
+size 94286

test_wavs/en/MDT_ASR_AF069A_A0006_S013_0_G0164_0077690-0078702.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fda02f33ab82c0c774cb71ed60b3d3f023ce9f7d32c17781b4a7ccb8b5ed0a0
+size 32846

test_wavs/en/MDT_ASR_AF069A_A0006_S013_0_G0164_0192562-0193377.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7913d796f86084419cc36e3ab6a62db144485b202c69914a0d860c2d9e1d448
+size 24654

test_wavs/en/MDT_ASR_AF069A_A0006_S013_0_G0164_0205010-0210417.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38ec020e3c2da65acd2da31e02450d4cac75cbaccbba8beac047339078b5b6f7
+size 173068

test_wavs/en/MDT_ASR_AF069A_A0006_S013_0_G0164_0237502-0238124.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7fe2858223d3fa170e0039abc22937125a8c36c752249730575391d77b9ed49
+size 19948

test_wavs/en/MDT_ASR_AF069A_A0030_S005_0_G0029_341940-342770.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa117eb6ca0e7d737b8b8a7555d41be18f8730bfd16ca5a1f3d707d7803c51af
+size 26604

test_wavs/en/MDT_ASR_AF069A_A0030_S007_0_G0024_653730-656780.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2446b0bf3972aafc8ac0bb991017604bb660a747c3cee771f0cd8df0c782cd41
+size 97644

test_wavs/en/MDT_ASR_AF069A_A0030_S007_0_G0024_783330-785120.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cecf9ef40496088c0c3c95f82d035f24d52a95e05bf315a017ab439ceceb8b08
+size 57422

test_wavs/en/MDT_ASR_AF069A_A0030_S007_0_G0024_860490-861890.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:516d4557fc5b0282031790055f4d3d4bfc45b81c52b6950422b2b8fdbb6cbe50
+size 45134