Upload 3 files

README.md

@@ -0,0 +1,60 @@
+# Baseline Orpheus TTS Combined Dataset
+
+## Overview
+This dataset contains synthetic speech generated using the unmodified Orpheus-3B model across all pathological and healthy speakers for baseline comparison.
+
+## Statistics
+- **Total Samples**: 800
+- **Total Duration**: 1534.04 seconds (0.4 hours)
+- **Average Duration**: 1.92 seconds
+- **Number of Speakers**: 8
+- **Sample Rate**: 24,000 Hz
+
+## Corpus Breakdown
+- **LibriSpeech_Healthy**: 60 samples, 554.4s total, 9.24s avg
+- **UA-Speech_Dysarthric**: 400 samples, 482.2s total, 1.21s avg
+- **TORGO_Healthy**: 200 samples, 289.8s total, 1.45s avg
+- **TORGO_Dysarthric**: 140 samples, 207.6s total, 1.48s avg
+
+
+## Speaker Coverage
+- **TORGO Dysarthric**: F04, M02
+- **TORGO Healthy**: FC02, MC01  
+- **UA-Speech**: F02, M04
+- **LibriSpeech**: 211, 4014
+
+## Model Information
+- **Base Model**: unsloth/orpheus-3b-0.1-ft (unmodified)
+- **Purpose**: Baseline comparison for pathological speech synthesis evaluation
+- **Generation Date**: 2025-09-13T14:36:30.220298
+
+## Usage
+```python
+from datasets import Dataset
+
+# Load combined dataset
+dataset = Dataset.from_parquet("baseline_orpheus_combined.parquet")
+
+# Filter by corpus
+torgo_samples = dataset.filter(lambda x: x['corpus'] == 'TORGO')
+dysarthric_samples = dataset.filter(lambda x: x['condition'] == 'Dysarthric')
+
+# Access audio and text
+for sample in dataset.select(range(3)):
+    print(f"Speaker {sample['speaker_id']}: {sample['text']}")
+    audio_array = sample['audio']['array']
+    print(f"Audio shape: {audio_array.shape}")
+```
+
+## Research Purpose
+This baseline dataset enables:
+- Comparison with fine-tuned pathological speech models
+- Evaluation of TTS adaptation effectiveness  
+- Baseline performance measurement for ASR degradation studies
+- Control condition for synthetic vs. real audio comparisons
+
+## Citation
+Please cite the original pathological speech synthesis paper and acknowledge:
+- Orpheus-3B model (unsloth/orpheus-3b-0.1-ft)
+- Original pathological speech corpora (TORGO, UA-Speech)
+- LibriSpeech corpus for healthy speech reference

baseline_orpheus_combined.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dccd59e2bc010f79675b1d1d0053e48de137c482e18e728369118e40335379d
+size 59121595

dataset_info.json

@@ -0,0 +1,101 @@
+{
+  "dataset_name": "baseline_orpheus_combined",
+  "total_samples": 800,
+  "statistics": {
+    "total_samples": 800,
+    "total_duration": 1534.0373333333293,
+    "avg_duration": 1.9175466666666665,
+    "corpus_breakdown": {
+      "LibriSpeech_Healthy": {
+        "count": 60,
+        "duration": 554.4106666666665
+      },
+      "UA-Speech_Dysarthric": {
+        "count": 400,
+        "duration": 482.21866666666625
+      },
+      "TORGO_Healthy": {
+        "count": 200,
+        "duration": 289.79200000000014
+      },
+      "TORGO_Dysarthric": {
+        "count": 140,
+        "duration": 207.6160000000001
+      }
+    },
+    "speakers": 8
+  },
+  "conversion_timestamp": "2025-09-14T11:33:13.136706",
+  "generation_summary": {
+    "generation_type": "baseline_orpheus_tts_for_comparison",
+    "total_datasets_processed": 8,
+    "total_baseline_audio_files": 800,
+    "generation_timestamp": "2025-09-13T14:36:30.220298",
+    "output_directory": "/content/drive/MyDrive/1training_stuff/baseline_orpheus_tts_20250913_125657",
+    "base_model": "unsloth/orpheus-3b-0.1-ft",
+    "model_type": "baseline_orpheus_3b",
+    "processing_results": {
+      "torgo_dysarthric_female": {
+        "files_generated": 70,
+        "status": "Success",
+        "dataset_path": "resproj007/torgo_dysarthric_female"
+      },
+      "torgo_dysarthric_male": {
+        "files_generated": 70,
+        "status": "Success",
+        "dataset_path": "resproj007/torgo_dysarthric_male"
+      },
+      "torgo_healthy_female": {
+        "files_generated": 100,
+        "status": "Success",
+        "dataset_path": "resproj007/torgo_healthy_female"
+      },
+      "torgo_healthy_male": {
+        "files_generated": 100,
+        "status": "Success",
+        "dataset_path": "resproj007/torgo_healthy_male"
+      },
+      "uaspeech_female": {
+        "files_generated": 200,
+        "status": "Success",
+        "dataset_path": "resproj007/uaspeech_female"
+      },
+      "uaspeech_male": {
+        "files_generated": 200,
+        "status": "Success",
+        "dataset_path": "resproj007/uaspeech_male"
+      },
+      "librispeech_female": {
+        "files_generated": 30,
+        "status": "Success",
+        "dataset_path": "resproj007/librispeech_female"
+      },
+      "librispeech_male": {
+        "files_generated": 30,
+        "status": "Success",
+        "dataset_path": "resproj007/librispeech_male"
+      }
+    },
+    "research_purpose": "Baseline comparison for pathological speech synthesis evaluation",
+    "comparison_datasets": [
+      "torgo_dysarthric_female",
+      "torgo_dysarthric_male",
+      "torgo_healthy_female",
+      "torgo_healthy_male",
+      "uaspeech_female",
+      "uaspeech_male",
+      "librispeech_female",
+      "librispeech_male"
+    ]
+  },
+  "individual_datasets": [
+    "librispeech_male_baseline",
+    "uaspeech_female_baseline",
+    "torgo_healthy_male_baseline",
+    "torgo_healthy_female_baseline",
+    "librispeech_female_baseline",
+    "torgo_dysarthric_female_baseline",
+    "torgo_dysarthric_male_baseline",
+    "uaspeech_male_baseline"
+  ]
+}