greenland

# GREENLAND Dataset Structure

## Dataset Organization

| Experiment Type | Category | Split Type | Subset | Files |
|----------------|----------|------------|---------|-------|
| Cross-lingual | Genetics/Script/Word Order | head_to_tail | major_to_major | train.csv, valid.csv, test.csv |
| | | | major_to_minor | train.csv, valid.csv, test.csv |
| | | | minor_to_major | train.csv, valid.csv, test.csv |
| | | | minor_to_minor | train.csv, valid.csv, test.csv |
| | | tail_to_tail | major_tail_to_minor_tail | train.csv, valid.csv, test.csv |
| | | | minor_tail_to_major_tail | train.csv, valid.csv, test.csv |
| | | all_to_all | all_head_to_all_tail | train.csv, valid.csv, test.csv |
| Multilingual | Genetics/Script/Word Order | head_and_head | major_head_to_major_head | train.csv, valid.csv, test.csv |
| | | | minor_head_to_minor_head | train.csv, valid.csv, test.csv |
| | | tail_and_tail | major_tail_to_major_tail | train.csv, valid.csv, test.csv |
| | | | minor_tail_to_minor_tail | train.csv, valid.csv, test.csv |
| | | all_and_all | all_head_to_all_head | train.csv, valid.csv, test.csv |
| | | | all_tail_to_all_tail | train.csv, valid.csv, test.csv |
| | | | head_and_tail_combined | train.csv, valid.csv, test.csv |

## Directory Structure

### Cross-lingual Experiments

cross_lingual/ ├── genetics/ │ ├── head_to_tail/ │ │ ├── major_to_major/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ ├── major_to_minor/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ ├── minor_to_major/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ └── minor_to_minor/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ ├── tail_to_tail/ │ │ ├── major_tail_to_minor_tail/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ └── minor_tail_to_major_tail/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ └── all_to_all/ │ └── all_head_to_all_tail/ │ ├── test.csv │ ├── train.csv │ └── valid.csv ├── script/ # Same structure as genetics └── word_order/ # Same structure as genetics

### Multilingual Experiments

multilingual/ ├── genetics/ │ ├── head_and_head/ │ │ ├── major_head_to_major_head/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ └── minor_head_to_minor_head/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ ├── tail_and_tail/ │ │ ├── major_tail_to_major_tail/ │ │ │ ├── test.csv │ │ │ ├── train.csv │ │ │ └── valid.csv │ │ └── minor_tail_to_minor_tail/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ └── all_and_all/ │ ├── all_head_to_all_head/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ ├── all_tail_to_all_tail/ │ │ ├── test.csv │ │ ├── train.csv │ │ └── valid.csv │ └── head_and_tail_combined/ │ ├── test.csv │ ├── train.csv │ └── valid.csv ├── script/ # Same structure as genetics └── word_order/ # Same structure as genetics

## Data Split Details

| Split Type | Description |
|------------|-------------|
| train.csv | Training dataset (80% for multilingual, 90% for cross-lingual) |
| valid.csv | Validation dataset (10%) |
| test.csv | Test dataset (10% for multilingual, 100% of target languages for cross-lingual) |

## Category Descriptions

| Category | Description |
|----------|-------------|
| Genetics | Language family-based splits |
| Script | Writing system-based splits |
| Word Order | Syntactic structure-based splits |

Would you like me to modify any part of this README structure?

Each category (genetics, script, word_order) follows identical internal structure but with different language splits based on their respective criteria:

1. Genetic Splits: Based on language families
2. Script Splits: Based on writing systems
3. Word Order Splits: Based on syntactic structure

Every terminal directory contains:

• train.csv: Training dataset
• valid.csv: Validation dataset
• test.csv: Test dataset

## File Structure Notes

Each experiment type (cross-lingual and multilingual) contains three main categories:
1. **Genetics**: Language family-based splits
2. **Script**: Writing system-based splits
3. **Word Order**: Syntactic structure-based splits

### Cross-lingual Splits
- **head_to_tail/**: Testing transfer from high-resource to low-resource languages
- **tail_to_tail/**: Testing transfer between low-resource languages
- **all_to_all/**: Comprehensive cross-resource evaluation

### Multilingual Splits
- **head_and_head/**: Transfer between high-resource languages
- **tail_and_tail/**: Transfer between low-resource languages
- **all_and_all/**: Combined resource level transfers

### Data Files
Each split contains three files:
- `train.csv`: Training data
- `valid.csv`: Validation data
- `test.csv`: Test data