mjbommar/opengloss-v1.2-contrastive-examples

---
license: cc-by-4.0
task_categories:
- text-generation
- sentence-similarity
- feature-extraction
language:
- en
tags:
- contrastive-learning
- semantic-similarity
- lexicon
- synonym
- antonym
- gradient
- synthetic
- education
- opengloss
size_categories:
- 10K<n<100K
---

# OpenGloss Contrastive Examples v1.2

## Dataset Summary

**OpenGloss Contrastive Examples** is a synthetic dataset of graduated semantic variations
designed for contrastive learning and semantic similarity training. Each example contains
a source sentence and a 5-point semantic gradient showing how meaning shifts from
antonym to synonym poles.

This dataset is derived from the [OpenGloss](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-definitions)
encyclopedic dictionary, using example sentences and their lexical context to generate
semantically coherent variations.

### Key Statistics

- **73,153 contrastive examples**
- **53,347 unique words/phrases**
- **53,446 unique lexemes**
- **5 gradient points per example** (0.0, 0.25, 0.5, 0.75, 1.0)
- **365,765 total sentence variations**
- **Average source sentence length: 10.8 words**
- **Average gradient sentence length: 11.1 words**

### Gradient Structure

Each example contains a 5-point semantic gradient:

| Position | Semantic Pole | Description |
|----------|---------------|-------------|
| 0.0 | Antonym | Semantic opposite of the target word |
| 0.25 | Near-antonym | Closer to opposite meaning |
| 0.5 | Neutral | Middle ground, balanced meaning |
| 0.75 | Near-synonym | Closer to original meaning |
| 1.0 | Synonym | Original word or close synonym |

### POS Distribution

| Part of Speech | Count |
|----------------|-------|
| noun | 51,182 |
| adjective | 12,076 |
| verb | 8,086 |
| adverb | 984 |
| determiner | 263 |
| preposition | 253 |
| interjection | 175 |
| pronoun | 78 |
| conjunction | 40 |
| proper noun | 8 |
| particle | 5 |
| numeral | 2 |
| abbreviation | 1 |

## Loading the Dataset

```python
from datasets import load_dataset

# Load the full dataset
dataset = load_dataset("mjbommar/opengloss-v1.2-contrastive-examples")

# Access records
for record in dataset["train"]:
print(f"Word: {record['word']}")
print(f"Source: {record['source_sentence']}")
print(f"Antonym: {record['antonym_word']} → {record['antonym_sentence']}")
print(f"Synonym: {record['synonym_word']} → {record['synonym_sentence']}\n")
```

## Example Record

```python
{
"id": "logjam_1_1_1",
"word": "logjam",
"lexeme_id": "logjam",
"pos": "verb",
"source_sentence": "Funding pipelines logjam during policy transitions.",
"gradient": [
{"position": 0.0, "word": "flow", "sentence": "Funding pipelines flow smoothly during policy transitions."},
{"position": 0.25, "word": "continue", "sentence": "Funding pipelines continue during policy transitions."},
{"position": 0.5, "word": "slow", "sentence": "Funding pipelines slow during policy transitions."},
{"position": 0.75, "word": "stall", "sentence": "Funding pipelines stall during policy transitions."},
{"position": 1.0, "word": "logjam", "sentence": "Funding pipelines logjam during policy transitions."}
],
"antonym_word": "flow",
"antonym_sentence": "Funding pipelines flow smoothly during policy transitions.",
"near_antonym_word": "continue",
"near_antonym_sentence": "Funding pipelines continue during policy transitions.",
"neutral_word": "slow",
"neutral_sentence": "Funding pipelines slow during policy transitions.",
"near_synonym_word": "stall",
"near_synonym_sentence": "Funding pipelines stall during policy transitions.",
"synonym_word": "logjam",
"synonym_sentence": "Funding pipelines logjam during policy transitions."
}
```

## Use Cases

### Contrastive Learning
Train models to understand semantic gradients and fine-grained meaning differences:

```python
# Create contrastive pairs
pairs = []
for record in dataset["train"]:
# Pair antonym with synonym for maximum contrast
pairs.append((record["antonym_sentence"], record["synonym_sentence"], 0.0))
# Pair neutral with synonym for medium contrast
pairs.append((record["neutral_sentence"], record["synonym_sentence"], 0.5))
```

### Semantic Similarity
Fine-tune embedding models with graduated similarity labels:

```python
# Each gradient position maps to a similarity score
for record in dataset["train"]:
for point in record["gradient"]:
similarity = point["position"] # 0.0 to 1.0
sentence = point["sentence"]
```

### Data Augmentation
Use gradient variations for text augmentation while controlling semantic shift:

```python
# Get semantically similar alternatives (position >= 0.75)
def get_similar_sentences(record):
return [p["sentence"] for p in record["gradient"] if p["position"] >= 0.75]
```

## Citation

If you use this dataset in your research, please cite:

```bibtex
@misc{bommarito2025opengloss_contrastive,
title={OpenGloss Contrastive Examples: Graduated Semantic Variations for Contrastive Learning},
author={Bommarito, Michael J., II},
year={2025},
url={https://huggingface.co/datasets/mjbommar/opengloss-v1.2-contrastive-examples},
note={Dataset available under CC-BY 4.0}
}
```

## License

This dataset is released under **Creative Commons Attribution 4.0 International (CC-BY 4.0)**.

## Related Datasets

- [OpenGloss v1.2 Dictionary](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-dictionary) - Word-level records
- [OpenGloss v1.2 Definitions](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-definitions) - Definition-level records
- [OpenGloss v1.2 Query Examples](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-query-examples) - Query-side retrieval supervision
- [OpenGloss v1.2 Hard Negative Pairs](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-hard-negative-pairs) - Calibration pairs for embedding training

## Acknowledgments

This dataset was generated using:
- [OpenGloss](https://huggingface.co/datasets/mjbommar/opengloss-v1.2-definitions) lexicon data
- OpenAI GPT models for gradient generation
- [pydantic-ai](https://github.com/pydantic/pydantic-ai) for structured generation

---

*Generated from the OpenGloss v1.2 lexicon.*