README.md

---
license: mit
language:
- en
---
# Laser Cleaning Query Relevance Dataset

## Overview

This dataset was created for training text classification models to identify customer queries relevant to laser cleaning services. It contains a comprehensive collection of text examples labeled for relevance to laser cleaning, enabling automated triage of customer inquiries for laser cleaning businesses.

## Dataset Description

- **Task**: Binary classification of text queries
- **Classes**: 
  - `1` = Relevant to laser cleaning
  - `0` = Not relevant to laser cleaning
- **Size**: Approximately 714 examples
  - Training set: 571 examples (80%)
  - Test set: 143 examples (20%)
- **Class distribution**:
  - Relevant queries: ~78% (557 examples)
  - Non-relevant queries: ~22% (157 examples)
- **Text length**: Short to medium queries (typically 5-25 words)
- **Languages**: English

## Files

The dataset is available in multiple formats:

- `full_dataset.jsonl` - Complete dataset in JSONL format
- `train.jsonl` - Training split in JSONL format
- `test.jsonl` - Testing split in JSONL format
- `train.csv` - Training split in CSV format
- `test.csv` - Testing split in CSV format

### File Format

#### JSONL Format
```json
{"text": "How does laser cleaning work?", "label": 1}
{"text": "Weather forecast for tomorrow", "label": 0}
```

#### CSV Format
```
text,label
"How does laser cleaning work?",1
"Weather forecast for tomorrow",0
```

## Dataset Creation

This dataset was systematically generated using multiple techniques:

1. **Base examples**: Manually curated positive examples (relevant to laser cleaning) and negative examples (not relevant).

2. **Template-based generation**: Using templates with placeholders to create numerous variations:
   ```
   "Can laser clean {thing}?"
   "Laser cleaning for {thing} - price?"
   ```
   
3. **Material and problem variations**: Systematic combination of materials (e.g., "car parts", "metal gate", "bronze statue") with cleaning problems (e.g., "rust", "corrosion", "paint").

4. **Compound questions**: Multi-faceted queries combining different aspects of laser cleaning.

5. **Paraphrasing**: Alternative phrasings of base examples to increase linguistic diversity.

6. **Ambiguous examples**: Carefully labeled edge cases to help models learn boundary conditions.

## Dataset Content

### Relevant Query Categories (Label 1)

- Specific laser cleaning questions
- Service area questions
- Quote and pricing related
- Material-specific inquiries
- Common applications
- Edge cases that are still relevant

Examples:
- "How does laser cleaning work?"
- "Do you offer laser cleaning in Huntsville?"
- "How much does laser cleaning cost?"
- "Laser cleaning for aluminum"
- "Using laser to clean motorcycle parts"
- "Rust removal options in Huntsville"

### Non-relevant Query Categories (Label 0)

- General information requests
- Other services entirely
- Random questions
- Consumer products
- Similarly worded but different domains
- Other industrial services

Examples:
- "Weather forecast for tomorrow"
- "Plumbing services in Huntsville"
- "What's the capital of France?"
- "Best laptop under $1000"
- "How to clean my computer keyboard"
- "CNC machining services"

## Usage

The dataset is formatted for easy use with common ML libraries:

### Loading with Pandas
```python
import pandas as pd

# Load CSV
train_df = pd.read_csv('train.csv')
test_df = pd.read_csv('test.csv')

# Access data
X_train = train_df['text'].values
y_train = train_df['label'].values
```

### Loading with PyTorch and Transformers
```python
from torch.utils.data import Dataset, DataLoader
from transformers import BertTokenizer

class LaserCleaningDataset(Dataset):
    def __init__(self, texts, labels, tokenizer, max_length=128):
        self.texts = texts
        self.labels = labels
        self.tokenizer = tokenizer
        self.max_length = max_length
        
    def __len__(self):
        return len(self.texts)
    
    def __getitem__(self, idx):
        text = str(self.texts[idx])
        label = self.labels[idx]
        
        # Tokenize the text
        encoding = self.tokenizer(
            text,
            add_special_tokens=True,
            max_length=self.max_length,
            return_token_type_ids=False,
            padding='max_length',
            truncation=True,
            return_attention_mask=True,
            return_tensors='pt'
        )
        
        return {
            'input_ids': encoding['input_ids'].flatten(),
            'attention_mask': encoding['attention_mask'].flatten(),
            'labels': torch.tensor(label, dtype=torch.long)
        }
```

## Applications

This dataset is designed for:

1. **Query triage**: Automatically identifying customer inquiries related to laser cleaning
2. **Chatbot development**: Training chatbots to recognize laser cleaning queries
3. **Customer service automation**: Routing queries to appropriate service representatives
4. **Search relevance**: Enhancing search functionality for laser cleaning businesses
5. **Marketing analysis**: Identifying potential customer needs related to laser cleaning

## License

This dataset is provided under the [MIT License](https://opensource.org/licenses/MIT), allowing for both academic and commercial use with minimal restrictions.

## Citation

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

```
RustBusters Laser Cleaning Query Relevance Dataset (2025)
Creator: RustBusters LLC
Version: 1.0
```

## Contact

For questions, improvements, or feedback about this dataset, please contact the RustBusters team.

---

*This dataset was carefully crafted to support machine learning applications in the laser cleaning industry.*