Upload folder using huggingface_hub

README.md

@@ -15,9 +15,11 @@ size_categories:
 - n<1K
 ---
 
-# Taiwan Mandarin–English Code-Switching Sentences (100)
+# TaiMECS: Taiwan Mandarin–English Code-Switching Sentences
 
-A set of 100 short sentences in **Taiwanese Mandarin with embedded English** (code-switching), spoken by a single Taiwanese speaker. Designed for training and evaluating **Taiwan-specific** Mandarin speech models, with deliberate emphasis on words whose **Taiwan readings differ from Mainland Mandarin**.
+A set of sentences in **Taiwanese Mandarin with embedded English** (code-switching), spoken by a single Taiwanese speaker. Designed for training and evaluating **Taiwan-specific** Mandarin speech models, with deliberate emphasis on words whose **Taiwan readings differ from Mainland Mandarin**.
+
+Pronunciations are **hand-curated and dictionary-verified** with a focus on hard, easy-to-get-wrong cases (Taiwan/Mainland differences, polyphonic characters, common model misreadings), making this a small but **high-quality** resource for Taiwan-specific speech. See [Pronunciation](#pronunciation).
 
 ## Composition
 
@@ -25,12 +27,15 @@ A set of 100 short sentences in **Taiwanese Mandarin with embedded English** (co
 | ----------- | ----------------------------------------------------------------------------------------------------------------------------- |
 | Clips       | 100 (`001.mp3`–`100.mp3`)                                                                                                     |
 | Audio       | MP3, 48 kHz, mono, 256 kbps                                                                                                   |
+| Loudness    | Normalized to **−16 LUFS**, true-peak **−1 dBTP** (EBU R128)                                                                 |
 | Duration    | 5–14 s each (~13 min total)                                                                                                   |
 | Speaker     | 1                                                                                                                             |
 | `human`     | 20 — recorded by the speaker                                                                                                  |
 | `synthetic` | 80 — generated with [openbmb/VoxCPM2](https://huggingface.co/openbmb/VoxCPM2), using a voice cloned from the human recordings |
 | Language    | Traditional Chinese (zh-Hant, Taiwan) + English                                                                               |
 
+All clips are loudness-normalized to **−16 LUFS** (integrated) with a **−1 dBTP** true-peak ceiling, using EBU R128 **two-pass linear** normalization (measure, then apply a single linear gain) so loudness is uniform across human and synthetic clips without altering dynamics.
+
 ## Format
 
 HuggingFace `AudioFolder` layout: audio files plus `metadata.jsonl`, with one line per clip:
@@ -53,9 +58,16 @@ HuggingFace `AudioFolder` layout: audio files plus `metadata.jsonl`, with one li
 
 ## Pronunciation
 
-The `text` field is the clean transcript/label.
+The `text` field is the clean transcript/label. The audio follows **Taiwan readings**, which were curated by hand with a deliberate focus on the *hardest* cases — situations where the correct Taiwan reading is easy to get wrong. This curation is what makes the dataset high-quality despite its small size.
+
+For the synthetic clips, Taiwan readings were enforced with **inline pinyin** (e.g., `企` → `{qi4}`; recorded in the `tts_text` field), covering four categories of difficult cases:
+
+- **Taiwan vs. Mainland readings** — characters whose standard Taiwan reading differs from Mainland Mandarin (e.g. 垃圾 *lè sè* not *lā jī*, 企業 *qì* not *qǐ*, 研究 *jiù* not *jiū*, conjunctive 和 *hàn* not *hé*).
+- **Polyphonic characters** — characters read differently by word sense, independent of the Taiwan/Mainland split (e.g. 還款 *huán* vs. 還是 *hái*, 調整 *tiáo* vs. 調度 *diào*, 成分 *fèn* vs. 分鐘 *fēn*).
+- **Model misreadings** — single-reading characters the strong TTS model still got wrong, surfaced through ASR comparison and spot listening, then forced to the correct reading (e.g. 房貸 *dài*, 崩潰 *kuì*, 肩頸 *jǐng*).
+- **Special tokens** — non-normalized tokens rewritten for correct delivery (e.g. `7-11` → "seven eleven").
 
-For synthesis, Taiwan readings were enforced with **inline pinyin** (e.g., `企` → `{qi4}`), covering three categories: Taiwan/Mainland pronunciation differences, polyphonic characters, and characters that the TTS system misread and were manually corrected. The synthetic voice’s soft retroflex realization of zh/ch/sh is an intended feature of the Taiwanese accent.
+Every reading above was cross-checked against the Taiwanese Ministry of Education dictionary (《重編國語辭典修訂本》). The synthetic voice’s soft retroflex realization of zh/ch/sh is an intended feature of the Taiwanese accent, not a defect.
 
 ## Intended Use
 

metadata.jsonl

@@ -24,7 +24,7 @@
 {"file_name": "024.mp3", "text": "寫論文最痛苦的就是 reference 的 format，每個期刊的要求都不一樣，光是排版就花掉我半天。", "source": "human", "tts_model": null, "reference_audio": null, "tts_text": null}
 {"file_name": "025.mp3", "text": "這個實驗的 sample size 不夠大，statistical significance 不足，審稿人一定會 challenge 這個結論。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "006.mp3", "tts_text": "這個實驗的 sample size 不夠大，statistical significance 不足，審稿人一定會 challenge 這個結論。"}
 {"file_name": "026.mp3", "text": "拜託你下樓的時候順便把 recycle 的垃圾拿去倒，今天剛好是回收日，錯過又要等一個禮拜。", "source": "human", "tts_model": null, "reference_audio": null, "tts_text": null}
-{"file_name": "027.mp3", "text": "我家附近新開了一間 gym，設備很新而且有教練帶 free 的體驗課，我打算這週末去 try 一下。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "003.mp3", "tts_text": "我家附近新開了一間 gym，設備很新而且有{jiao4}練帶 free 的體驗課，我打算這週{mo4}去 try 一下。"}
+{"file_name": "027.mp3", "text": "我家附近新開了一間 gym，設備很新而且有教練帶 free 的體驗課，我打算這週末去 try 一下。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "001.mp3", "tts_text": "我家附近新開了一間 gym，設備很新而且有{jiao4}練帶 free 的體驗課，我打算這{zhou1}{mo4}去 try 一下。"}
 {"file_name": "028.mp3", "text": "這支手機的 battery 真的很不耐用，明明早上才充飽，到下午就剩百分之二十，續航力超差。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "002.mp3", "tts_text": "這支手機的 battery 真的很不耐用，明明早上才充{bao3}，到下午就剩百分之二十，續航力超{cha4}。"}
 {"file_name": "029.mp3", "text": "我昨天網購的東西到了，可是 size 拿錯了，現在要走 return 的流程，超級麻煩。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "004.mp3", "tts_text": "我昨天網{gou4}的東西到了，可是 size 拿錯了，現在要走 return 的流程，超級麻煩。"}
 {"file_name": "030.mp3", "text": "最近物價漲得好兇，連去 7-11 買個早餐都覺得貴，薪水根本跟不上 inflation 的速度。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "005.mp3", "tts_text": "最近物價{zhang3}得好兇，連去 seven eleven 買個早餐都覺得貴，薪水根本跟不上 inflation 的速度。"}
@@ -50,11 +50,11 @@
 {"file_name": "050.mp3", "text": "加密貨幣的市場波動非常劇烈，又缺乏明確的 regulation，一般散戶很容易在這種市場裡受傷。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "009.mp3", "tts_text": "加密貨幣的市場波動非常劇烈，又缺乏明確的 regulation，一般{san3}戶很容易在這種市場裡受傷。"}
 {"file_name": "051.mp3", "text": "設計稿的 layout 我覺得還不錯，但 color palette 太暗了，整體質感看起來有點壓抑。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "010.mp3", "tts_text": "設計稿的 layout 我覺得還不錯，但 color palette 太暗了，整體{zhi2}感看起來有點壓抑。"}
 {"file_name": "052.mp3", "text": "客戶要的 logo 風格一直變來變去，我們改了不下十個 version，整個 design team 都快崩潰了。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "001.mp3", "tts_text": "客戶要的 logo 風格一直變來變去，我們改了不下十個 version，整個 design team 都快崩{kui4}了。"}
-{"file_name": "053.mp3", "text": "這支影片的 storytelling 很有渲染力，配上節奏明快的 BGM，難怪上線一週就破百萬觀看。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "012.mp3", "tts_text": "這支影片的 storytelling 很有渲染力，配上節奏明快的 BGM，難怪上線一週就破百萬觀看。"}
+{"file_name": "053.mp3", "text": "這支影片的 storytelling 很有渲染力，配上節奏明快的 BGM，難怪上線一週就破百萬觀看。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "002.mp3", "tts_text": "這支影片的 storytelling 很有渲染力，配上節奏明快的 BGM，難怪上線一{zhou1}就破百萬觀看。"}
 {"file_name": "054.mp3", "text": "UI 設計最重要的是 user flow 要直覺，如果使用者找不到關鍵按鈕，再漂亮的介面也沒意義。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "003.mp3", "tts_text": "UI 設計最重要的是 user flow 要{zhi2}{jue2}，如果使用者找不到關鍵按鈕，再漂亮的介面也沒意義。"}
 {"file_name": "055.mp3", "text": "這次品牌改版我們重新定義了 brand identity，從字體到配��都更新，希望傳達更年輕的形象。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "017.mp3", "tts_text": "這次品牌改版我們{chong2}新定義了 brand identity，從字體到配色都{geng1}新，希望傳達更年輕的形象。"}
 {"file_name": "056.mp3", "text": "我們這輪募資見了好幾個 VC，但他們對我們的 business model 還是有疑慮，估值一直談不攏。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "019.mp3", "tts_text": "我們這輪募資見了好幾個 VC，但他們對我們的 business model 還是有疑慮，估值一直談不攏。"}
-{"file_name": "057.mp3", "text": "創業最辛苦的就是 cash flow，明明訂單一直進來，但客戶帳期太長，公司隨時可能週轉不靈。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "001.mp3", "tts_text": "創業最辛苦的就是 cash flow，明明訂單一直進來，但客戶{zhang4}{qi2}太{chang2}，公司隨時可能週轉不靈。"}
+{"file_name": "057.mp3", "text": "創業最辛苦的就是 cash flow，明明訂單一直進來，但客戶帳期太長，公司隨時可能週轉不靈。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "003.mp3", "tts_text": "創業最辛苦的就是 cash flow，明明訂單一直進來，但客戶{zhang4}{qi2}太{chang2}，公司隨時可能{zhou1}轉不靈。"}
 {"file_name": "058.mp3", "text": "這家公司的企業文化很 open，員工有任何 idea 都可以直接跟 CEO 提，決策的速度非常快。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "021.mp3", "tts_text": "這家公司的{qi4}業文化很 open，員工有任何 idea 都可以直接跟 CEO 提，決策的速度非常快。"}
 {"file_name": "059.mp3", "text": "我們的產品終於找到 product market fit，使用者的 retention 明顯上升，對團隊是很大的鼓舞。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "023.mp3", "tts_text": "我們的產品終於找到 product market fit，使用者的 retention 明顯上升，對團隊是很大的鼓舞。"}
 {"file_name": "060.mp3", "text": "Pivot 之後我們把 target 客群從個人改成中小企業，沒想到市場的反應比預期好很多。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "024.mp3", "tts_text": "Pivot 之後我們把 target 客群從個人改成中小{qi4}業，沒想到市場的反{ying4}比預{qi2}好很多。"}
@@ -78,7 +78,7 @@
 {"file_name": "078.mp3", "text": "遠距工作雖然彈性，但有時候 communication 的成本反而更高，很多事情用 Slack 講不清楚。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "003.mp3", "tts_text": "{yuan3}距工作雖然{tan2}性，但有時候 communication 的成本反而更高，很多事情用 Slack 講不清楚。"}
 {"file_name": "079.mp3", "text": "我們公司最近在推 OKR，希望每個人的目標都能跟公司的願景 align，但執行起來其實滿卡的。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "007.mp3", "tts_text": "我們公司最近在推 OKR，希望每個人的目標都能跟公司的{yuan4}景 align，但執行起來其實滿卡的。"}
 {"file_name": "080.mp3", "text": "那位同事的 attitude 真的有問題，每次 collaborate 都把責任推給別人，team work 很難進行。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "024.mp3", "tts_text": "那位同事的 attitude 真的有問題，每次 collaborate 都把責任推給別人，team work 很難進行。"}
-{"file_name": "081.mp3", "text": "這部影集的 plot twist 真的太神了，我一個週末就把整季 binge 完，結局完全出乎意料。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "008.mp3", "tts_text": "這部影集的 plot twist 真的太神了，我一個週{mo4}就把整季 binge 完，結局完全出乎意料。"}
+{"file_name": "081.mp3", "text": "這部影集的 plot twist 真的太神了，我一個週末就把整季 binge 完，結局完全出乎意料。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "004.mp3", "tts_text": "這部影集的 plot twist 真的太神了，我一個{zhou1}{mo4}就把整季 binge 完，結局完全出乎意料。"}
 {"file_name": "082.mp3", "text": "我最近迷上手沖咖啡，光是研究水溫和粉水比的 ratio 就花了不少時間，但泡出來的味道真的差很多。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "001.mp3", "tts_text": "我最近迷上手沖咖啡，光是研{jiu4}水溫{han4}粉水比的 ratio 就花了不少時間，但泡出來的味道真的差很多。"}
 {"file_name": "083.mp3", "text": "這款掃地機器人的 sensor 很靈敏，會自動避開障礙物，連桌腳和電線都不太會卡住。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "001.mp3", "tts_text": "這{kuan3}{sao3}地機器人的 sensor 很靈敏，會自動避開障礙物，連桌腳{han4}電線都不太會{ka3}住。"}
 {"file_name": "084.mp3", "text": "颱風天最怕停電，建議大家隨身攜帶 power bank 和手電筒，把重要的東西都 charge 飽。", "source": "synthetic", "tts_model": "openbmb/VoxCPM2", "reference_audio": "008.mp3", "tts_text": "{tai2}風天最怕停電，建議大家隨身{xi1}帶 power bank {han4}手電筒，把重要的東西都 charge {bao3}。"}