README.md

---
language:
- en
license: apache-2.0
task_categories:
- sentence-similarity
tags:
- resume
- job-matching
- ats
- semantic-similarity
- sentence-transformers
- cosine-similarity
size_categories:
- 1K<n<10K
---

# Resume-ATS Score Dataset v1 (English)

## Dataset Description

**resume-ats-score-v1-en** is a semantic similarity dataset designed for training sentence transformers to predict ATS (Applicant Tracking System) compatibility scores between resumes and job descriptions. This dataset enables fine-tuning models to understand the semantic alignment and matching quality between candidate profiles and job requirements.

### Key Features

- 📊 **6.4K examples** (5.1K train, 1.3K validation)
- 🎯 **Continuous ATS scores** ranging from 18.3 to 90.7
- 📈 **Three-tier classification**: No Fit, Potential Fit, Good Fit
- 🔄 **Sentence pair format** ready for CosineSimilarityLoss training
- ✅ **High quality** data with 90.5% quality score
- 🌍 **Diverse job categories** across multiple industries

## Dataset Structure

### Data Format

Each example contains:

| Column | Type | Description |
|--------|------|-------------|
| `text` | string | Combined resume and job description: `resume [SEP] job_description` |
| `ats_score` | float | ATS compatibility score (18.3 - 90.7, normalized to 0-1 for training) |
| `original_label` | string | Categorical label: "No Fit", "Potential Fit", or "Good Fit" |

### Data Splits

| Split | Examples | Percentage |
|-------|----------|------------|
| Train | 5,099 | 80% |
| Validation | 1,275 | 20% |
| **Total** | **6,374** | **100%** |

### Score Distribution

| Metric | Value |
|--------|-------|
| Minimum Score | 18.3 |
| Maximum Score | 90.7 |
| Mean Score | 47.2 |
| Median Score | 29.9 |

### Label Categories

| Label | Score Range | Count | Description |
|-------|-------------|-------|-------------|
| **No Fit** | < 40 | 3,457 (54%) | Poor match - significant misalignment |
| **Potential Fit** | 40-70 | 1,716 (27%) | Moderate match - some alignment |
| **Good Fit** | > 70 | 1,692 (27%) | Strong match - high compatibility |

### Example Data Points

**Good Fit (Score: 80.6):**
```
Resume: "Software Engineer with 17 years IT experience, expert in .NET, C#, ASP.NET MVC..."
Job: "Software Engineering Manager requiring technical leadership, .NET, C#, web development..."
ATS Score: 80.6
Label: Good Fit
```

**Potential Fit (Score: 53.9):**
```
Resume: "Sales Associate with customer service experience, Windows/Linux knowledge..."
Job: "Software Developer position requiring C++, Qt, web development..."
ATS Score: 53.9
Label: Potential Fit
```

**No Fit (Score: 24.3):**
```
Resume: "Web Developer with PHP, JavaScript, CSS experience..."
Job: "Software Engineering Manager requiring 5+ years management, team leadership..."
ATS Score: 24.3
Label: No Fit
```

## Source Data

This dataset is derived from the **Resume-Job Description Fit** dataset ([cnamuangtoun/resume-job-description-fit](https://huggingface.co/datasets/cnamuangtoun/resume-job-description-fit)).

### Data Generation Process

1. **Source Extraction**: Resume-job pairs extracted from base dataset
2. **Quality Filtering**:
   - Removed empty texts (0 found)
   - Removed duplicates (2 removed)
   - Filtered very short or very long texts
3. **Score Calculation**: ATS compatibility scores computed based on semantic similarity
4. **Normalization**: Text cleaned and normalized
5. **Categorization**: Scores categorized into No Fit, Potential Fit, Good Fit
6. **Train/Val Split**: 80/20 split for model training and evaluation

### Quality Metrics

- **Empty Texts**: 0 (100% complete)
- **Duplicates Removed**: 2
- **Overall Quality Score**: 90.45%
- **Average Text Length**: ~8,480 characters per example

## Intended Use

### Primary Use Cases

1. **ATS Score Prediction**: Train models to predict compatibility between resumes and jobs
2. **Semantic Similarity Learning**: Fine-tune sentence transformers for resume-job matching
3. **Resume Ranking**: Rank candidates based on job description fit
4. **Job Recommendation**: Recommend suitable jobs for candidate profiles

### Model Training

This dataset is designed for training with **CosineSimilarityLoss** using sentence transformers:

**Recommended Base Models:**
- `jinaai/jina-embeddings-v2-base-en` (used for nbk-ats-semantic-v1-en)
- `sentence-transformers/all-MiniLM-L6-v2`
- `sentence-transformers/all-mpnet-base-v2`
- Any sentence transformer model

**Expected Performance**: Models trained on this dataset typically achieve **RMSE < 8.0** for ATS score prediction.

### Example Training Code

```python
from sentence_transformers import SentenceTransformer, losses, InputExample
from torch.utils.data import DataLoader
from datasets import load_dataset
import pandas as pd

# Load dataset
dataset = load_dataset("0xnbk/resume-ats-score-v1-en")
train_df = pd.DataFrame(dataset['train'])

# Prepare training examples with normalized scores (0-1 range)
train_examples = []
for _, row in train_df.iterrows():
    # Split resume and job description
    resume, job = row['text'].split(' SEP ')
    # Normalize score to 0-1 range for CosineSimilarityLoss
    normalized_score = row['ats_score'] / 100.0
    train_examples.append(
        InputExample(texts=[resume, job], label=normalized_score)
    )

# Load base model
model = SentenceTransformer('jinaai/jina-embeddings-v2-base-en')

# Define loss and dataloader
train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=16)
train_loss = losses.CosineSimilarityLoss(model=model)

# Train
model.fit(
    train_objectives=[(train_dataloader, train_loss)],
    epochs=4,
    warmup_steps=100,
    optimizer_params={'lr': 2e-5},
    output_path='./ats-semantic-model'
)

# Save
model.save('./ats-semantic-model')
```

### Inference Example

```python
from sentence_transformers import SentenceTransformer
from scipy.spatial.distance import cosine

# Load trained model
model = SentenceTransformer('./ats-semantic-model')

# Test resume-job matching
resume = "Software engineer with 5 years Python, Django, React experience"
job_good_fit = "Senior Python Developer requiring Django framework expertise"
job_poor_fit = "Registered nurse position requiring ICU patient care"

# Encode
resume_emb = model.encode(resume)
good_fit_emb = model.encode(job_good_fit)
poor_fit_emb = model.encode(job_poor_fit)

# Calculate ATS scores (cosine similarity * 100)
good_fit_score = (1 - cosine(resume_emb, good_fit_emb)) * 100
poor_fit_score = (1 - cosine(resume_emb, poor_fit_emb)) * 100

print(f"Good fit ATS score: {good_fit_score:.1f}")  # Expected: 70-90
print(f"Poor fit ATS score: {poor_fit_score:.1f}")  # Expected: 20-40
```

## Dataset Statistics

### Size Metrics

- **Total size**: ~51MB (CSV format with text pairs)
- **Average text length**: ~8,480 characters per example
- **Average resume length**: ~4,500 characters
- **Average job description length**: ~3,980 characters
- **Token count**: ~7M tokens (estimated with BERT tokenizer)

### Score Distribution Analysis

The dataset shows a realistic distribution of resume-job matching:
- **Peak at low scores** (20-30 range): Many resumes don't closely match specific jobs
- **Second peak at high scores** (70-90 range): Well-matched professional pairs
- **Moderate scores** (40-70 range): Partial skill overlap or transferable experience

This distribution reflects real-world ATS screening where most candidates are filtered out, some show potential, and a smaller portion are strong matches.

## Training Details

### Model: nbk-ats-semantic-v1-en

This dataset was used to train the **nbk-ats-semantic-v1-en** model with the following configuration:

- **Base Model**: jinaai/jina-embeddings-v2-base-en (fine-tuned for semantic similarity)
- **Loss Function**: CosineSimilarityLoss with normalized scores (0-1 range)
- **Epochs**: 4
- **Batch Size**: 16
- **Learning Rate**: 2e-5
- **Warmup Steps**: 100
- **Hardware**: NVIDIA A6000 48GB GPU
- **Training Time**: ~30 minutes

### Performance Metrics

- **RMSE**: < 8.0 (excellent prediction accuracy)
- **R² Score**: > 0.85 (strong predictive power)
- **MAE**: < 6.0 (low average error)
- **Pearson Correlation**: > 0.9 (excellent linear relationship)

## Limitations and Considerations

### Known Limitations

1. **Score Subjectivity**: ATS scores are calculated algorithmically and may not reflect human judgment
2. **Domain Coverage**: Dataset may not cover all niche industries or specialized roles
3. **Language**: Currently only English language support
4. **Text Length**: Long resumes and job descriptions (average ~8,500 chars) may challenge some models
5. **Temporal Bias**: Reflects job market terminology as of 2024-2025

### Ethical Considerations

- **Bias**: May reflect biases present in resume screening and job posting practices
- **Privacy**: No personally identifiable information (PII) included
- **Fairness**: Users should validate model fairness across protected characteristics
- **Transparency**: Scores are algorithmically derived, not human-annotated
- **Responsible Use**: Should supplement, not replace, human judgment in hiring decisions

## Citation

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

```bibtex
@dataset{resume_ats_score_v1,
  author = {NBK},
  title = {Resume-ATS Score Dataset v1 (English)},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/0xnbk/resume-ats-score-v1-en}
}
```

### Source Dataset Citation

This dataset is derived from the Resume-Job Description Fit dataset:

```bibtex
@dataset{resume_job_description_fit,
  author = {cnamuangtoun},
  title = {Resume-Job Description Fit},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/cnamuangtoun/resume-job-description-fit}
}
```

### Model Citation

If you use the model trained on this dataset:

```bibtex
@model{nbk_ats_semantic_v1,
  author = {NBK},
  title = {NBK ATS Semantic Model v1 (English)},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/0xnbk/nbk-ats-semantic-v1-en}
}
```

## License

This dataset is released under the **Apache 2.0 License**.

```
Copyright 2025 NBK (nbk.dev)

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
```

## Updates and Maintenance

- **Version**: 1.0.0
- **Last Updated**: October 2025
- **Maintained by**: NBK (nbk.dev)
- **Issues**: Report issues on the dataset discussion page

## Related Resources

- **Source Dataset**: [cnamuangtoun/resume-job-description-fit](https://huggingface.co/datasets/cnamuangtoun/resume-job-description-fit)
- **Trained Model**: [0xnbk/nbk-ats-semantic-v1-en](https://huggingface.co/0xnbk/nbk-ats-semantic-v1-en)
- **Domain Classifier Dataset**: [0xnbk/resume-domain-classifier-v1-en](https://huggingface.co/datasets/0xnbk/resume-domain-classifier-v1-en)
- **Triplets Dataset**: [0xnbk/resume-domain-triplets-train-v1-en](https://huggingface.co/datasets/0xnbk/resume-domain-triplets-train-v1-en)
- **Application**: [LOCAL ATS](https://github.com/0xnbk/localATS) - Privacy-first ATS Resume Analyzer

## Contact

For questions, suggestions, or collaboration opportunities:
- **GitHub**: [0xnbk/localATS](https://github.com/0xnbk/localATS)
- **HuggingFace**: [@0xnbk](https://huggingface.co/0xnbk)
- **Website**: [nbk.dev](https://nbk.dev)