Update README.md

README.md

@@ -2,61 +2,95 @@
 language:
 - zh
 - en
-license: cc-by-nc-4.0
-task_categories:
-- automatic-speech-recognition
 tags:
+- automatic-speech-recognition
 - code-switching
+- audio
+- speech-processing
+license: cc-by-nc-sa-4.0
+task_categories:
+- automatic-speech-recognition
 dataset_info:
-  config_names:
-  - SECoMiCSC
-  - DevCECoMiCSC
   features:
-    - name: file_name
-      dtype: string
-    - name: sentence
-      dtype: string
-    - name: duration
-      dtype: float32
-    - name: source
-      dtype: string
-splits:
-  - name: train
-    num_bytes: 0
-    num_rows: 0
-  - name: test
-    num_bytes: 0
-    num_rows: 0
+  - name: audio
+    dtype: audio
+  - name: sentence
+    dtype: string
+  - name: duration
+    dtype: float32
+  - name: source
+    dtype: string
+  - name: original_tag
+    dtype: string
 ---
 
-# Robust Code-Switching ASR Benchmark
+# Unified Mandarin-English Code-Switching Dataset (Processed)
+
+<div align="center">
+    <img src="https://huggingface.co/front/assets/huggingface_logo-noborder.svg" width="50" height="50"/>
+</div>
 
-## Dataset Summary
-This dataset is a **processed and cleaned derivative** of the open-source MagicData corpus, specifically optimized for our project **Code-Switched ASR robustness** (e.g., Whisper fine-tuning). 
+This dataset is a curated compilation of **[SECoMiCSC](https://magichub.com/datasets/chinese-english-code-mixing-conversational-speech-corpus/)**, **[DevCECoMiCSC](https://magichub.com/datasets/dev-set-of-chinese-english-code-mixing-conversational-speech-corpus/)**, and **[BAAI/CS-Dialogue](https://huggingface.co/datasets/BAAI/CS-Dialogue)**, specifically processed for Code-Switching ASR research.
 
-We addressed the "context fragmentation" issue in original long-form audio by applying a **Smart-Merge Strategy** (merging short segments into 5-15s chunks using ground-truth timestamps) and filtering out conversational fillers.
+```text
+root/
+├── audio/
+│   ├── SECoMiCSC/        # Chunked segments from SECoMiCSC
+│   ├── DevCECoMiCSC/     # Chunked segments from DevCECoMiCSC
+│   └── CS_Dialogue/      # Extracted <MIX> segments from BAAI/CS-Dialogue
+├── metadata.jsonl        # Universal index containing paths, transcripts, and metadata
+└── data_preparation.py   # Script to reproduce this dataset from raw sources
+```
 
-## Original Data Sources
-This dataset is derived from the following open-source datasets released by **MagicData Technology**:
+## Usage
 
-* **Training Subset:** Derived from **ASR-SECoMiCSC**
-    * *Source:* [MagicData Open Source Community](https://magichub.com/datasets/chinese-english-code-mixing-conversational-speech-corpus/)
-* **Benchmark/Test Subset:** Derived from **ASR-DevCECoMiCSC**
-    * *Source:* [MagicData Open Source Community](https://magichub.com/datasets/dev-set-of-chinese-english-code-mixing-conversational-speech-corpus/)
+```python
+from datasets import load_dataset, Audio
 
-*> Note: This repository contains processed audio chunks and metadata only. Please refer to the original links for full datasets and license details.*
+# Load with streaming (Recommended)
+data = load_dataset("1uckyan/code-switch_chunks", split="train", streaming=True)
 
-## Processing Pipeline (Why this version?)
-1.  **Smart Segmentation:** Instead of random VAD cutting, we merged short utterances into **5s - 15s segments** based on speaker identity and time gaps. This provides better context for Transformer-based models.
-2.  **Noise Filtering:** Removed pure filler segments (e.g., "嗯", "啊", "[ENS]") to reduce hallucination during training.
+# Important: Cast to 16kHz
+data = data.cast_column("audio", Audio(sampling_rate=16000))
 
-## Usage
+for sample in data:
+    print(f"Source: {sample['source']} | Text: {sample['sentence']}")
+    break
+```
 
-```python
-from datasets import load_dataset
 
-# 1. Load Training Data (SECoMiCSC)
-dataset_train = load_dataset("1uckyan/code-switch_chunks", data_dir="SECoMiCSC", split="train")
+## Data Sources & Creation
+
+
+| Source Dataset | Original Content | Processing / Cleaning Logic |
+| --- | --- | --- |
+| **SECoMiCSC** | Conversational Speech | **VAD-based Chunking**: Split >1.8s gaps, merged to 5-15s segments.
+| **DevCECoMiCSC** | Conversational Speech | **VAD-based Chunking**: Same as above. |
+| **BAAI/CS-Dialogue** | Dialogue | **Tag Filtering**: Only retained utterances tagged as <MIX>
+
+
+
+## Reproducibility
+
+We provide the `data_preparation.py` script in this repository to ensure the transparency and reproducibility of our data processing pipeline.
+
+If you have access to the raw source datasets, you can recreate this specific processed version by running:
+
+```bash
+python data_preparation.py \
+  --secomicsc_root /path/to/local/ASR-SECoMiCSC \
+  --dev_root /path/to/local/ASR-DevCECoMiCSC \
+  --cs_dialogue_root /path/to/local/CS_Dialogue/data/short_wav \
+  --output_dir ./output_Dataset
+
+```
+
+## License & Citations
+
+This dataset is a derivative work. We adhere to the licenses of the original source datasets:
+
+* **BAAI/CS-Dialogue**: Licensed under **CC BY-NC-SA 4.0**.
+* **SECoMiCSC / DevCECoMiCSC**: Please refer to their original publications for usage rights.
+
+If you use this dataset, please cite the original authors of the source datasets and our work.
 
-# 2. Load Benchmark Test Set (DevCECoMiCSC)
-dataset_test = load_dataset("1uckyan/code-switch_chunks", data_dir="DevCECoMiCSC", split="train")