Update README.md

README.md

@@ -1,3 +1,183 @@
----
-license: mit
----
+---
+license: mit
+---
+# RustBusters Laser Cleaning QA Dataset
+
+The RustBusters Laser Cleaning QA dataset contains 3,000 synthetic question-answer pairs designed for training a customer service assistant for RustBustersHSV, a laser cleaning and resurfacing company in Huntsville, Alabama.
+
+## Dataset Description
+
+- **Repository:** https://huggingface.co/datasets/RustBustersHSV/laser-cleaning-qa
+- **Language:** English
+- **Format:** JSONL (JSON Lines)
+- **Size:** 3,000 QA pairs
+- **Creation Date:** 2025
+- **License:** MIT
+
+### Dataset Summary
+
+This dataset consists of synthetically generated question-answer pairs designed to train a customer service assistant for a laser cleaning business. The questions cover various aspects of laser cleaning services, from general inquiries to technical specifications. The answers are crafted to be warm, informative, and to guide potential customers toward requesting a free quote.
+
+## Dataset Creation
+
+### Curation Rationale
+
+The dataset was created to fine-tune a language model for specialized customer service interactions in the laser cleaning industry. The goal was to generate a comprehensive set of questions and answers that cover the most common customer inquiries, concerns, and information needs related to laser cleaning services.
+
+### Source Data
+
+The dataset was synthetically generated using template-based techniques. No real customer data was used in its creation, ensuring privacy and confidentiality.
+
+#### Initial Data Collection and Normalization
+
+Base prompts were manually created for seven key categories relevant to laser cleaning services:
+
+1. General inquiries
+2. Service-specific questions
+3. Logistics and location information
+4. Process details
+5. Concerns and objections
+6. Customer experience
+7. Technical aspects
+
+These base prompts were then expanded using:
+- Various prefix templates (e.g., "Tell me about", "I want to know more about")
+- Suffix variations (e.g., "?", " in detail?", " for my project?")
+- Category-specific response templates with variable slots
+
+#### Who Generated the Data
+
+The data was programmatically generated using the template approach described above, with templates designed by RustBustersHSV staff to accurately represent their services and customer interactions.
+
+### Data Structure and Format
+
+The dataset is provided in JSONL (JSON Lines) format, with each line containing a single QA pair with metadata:
+
+```json
+{
+  "instruction": "What is laser cleaning and how does it work?",
+  "response": "Laser cleaning uses focused light energy to remove contaminants from surfaces without chemicals or abrasives. The laser vaporizes unwanted materials while preserving the base material. It's precise, environmentally friendly, and doesn't damage the underlying material. We'd be happy to provide a free quote for your specific project!",
+  "metadata": {
+    "source_prompt": "What is laser cleaning and how does it work?",
+    "category": "general_inquiries"
+  }
+}
+```
+
+#### Data Fields
+
+- **instruction**: The customer question or query
+- **response**: The assistant's answer to the question
+- **metadata**: Additional information about the QA pair
+  - **source_prompt**: The original prompt used to generate this QA pair
+  - **category**: The category of the question (e.g., "general_inquiries", "service_specific", etc.)
+
+### Data Statistics
+
+The dataset contains 3,000 QA pairs distributed across seven categories:
+
+| Category | Approximate Count | Description |
+|----------|------------------|-------------|
+| general_inquiries | ~450 | Basic questions about laser cleaning |
+| service_specific | ~450 | Questions about specific cleaning services |
+| logistics_location | ~450 | Questions about location, hours, and service area |
+| process_details | ~450 | Questions about the laser cleaning process |
+| concerns_objections | ~450 | Questions addressing common concerns |
+| customer_experience | ~450 | Questions about what to expect as a customer |
+| technical_aspects | ~300 | Technical questions about laser technology |
+
+#### Example Entries
+
+**Example 1: General Inquiry**
+```json
+{
+  "instruction": "What is laser cleaning and how does it work?",
+  "response": "Laser cleaning uses focused light energy to remove rust from surfaces without chemicals or abrasives. The laser vaporizes unwanted materials while preserving the base material. It's precise, environmentally friendly, and suitable for delicate items. We'd be happy to provide a free quote for your specific cleaning needs!",
+  "metadata": {
+    "source_prompt": "What is laser cleaning and how does it work?",
+    "category": "general_inquiries"
+  }
+}
+```
+
+**Example 2: Service Specific**
+```json
+{
+  "instruction": "How long does laser cleaning typically take for a car wheel?",
+  "response": "For automotive parts, our laser cleaning process is particularly effective. It removes the rust while preserving the base material. The process typically takes as little as 30 minutes depending on the size and condition. We'd be happy to assess your specific needs and provide a personalized quote!",
+  "metadata": {
+    "source_prompt": "How long does laser cleaning typically take for a car wheel?",
+    "category": "service_specific"
+  }
+}
+```
+
+**Example 3: Technical Aspect**
+```json
+{
+  "instruction": "What type of laser do you use for cleaning?",
+  "response": "Our fiber laser system operates at specified wavelengths, which allows for selective material removal. This technology eliminates the need for chemicals or abrasives compared to conventional methods. It's particularly effective for automotive restoration because of its precision. We'd be happy to discuss the technical specifications during a free consultation!",
+  "metadata": {
+    "source_prompt": "What type of laser do you use for cleaning?",
+    "category": "technical_aspects"
+  }
+}
+```
+
+## Intended Uses
+
+This dataset is primarily intended for:
+
+1. Fine-tuning language models to serve as customer service assistants for laser cleaning businesses
+2. Training conversational agents specializing in industrial cleaning services
+3. Educational purposes for understanding customer service interactions in specialized technical fields
+4. Research into domain-specific language adaptation
+
+### Considerations for Using the Data
+
+- The dataset is specifically designed for the laser cleaning industry and may not transfer well to other domains
+- All responses guide customers toward requesting quotes, which may not be appropriate for all use cases
+- The synthetic nature of the data means it may not capture the full diversity of real customer inquiries
+- The responses present laser cleaning in a positive light and may not adequately address all potential drawbacks
+
+## Dataset Creation Method
+
+The dataset was created using a template-based approach with randomized variations:
+
+1. **Base Templates**: Each category had several core question templates and response templates
+2. **Response Components**: Variable components like "benefit1", "contaminant", or "process_detail1" had multiple possible values
+3. **Template Filling**: For each QA pair, a template was selected and variables were filled with random but contextually appropriate values
+4. **Prefix/Suffix Variations**: Questions were further diversified using prefixes and suffixes
+
+Example template:
+```
+"Laser cleaning uses focused light energy to remove {contaminant} from surfaces without chemicals or abrasives. The laser {action} unwanted materials while preserving the base material. It's {benefit1}, {benefit2}, and {benefit3}. We'd be happy to provide a free quote for your specific cleaning needs!"
+```
+
+Where variables like `{contaminant}` would be replaced with values from predefined lists (e.g., "rust", "paint", "oxidation").
+
+## Dataset Citation
+
+```
+@misc{rustbusters_laser_cleaning_dataset,
+  title = {RustBusters Laser Cleaning QA Dataset},
+  author = {RustBustersHSV},
+  year = {2025},
+  publisher = {Hugging Face},
+  journal = {Hugging Face Dataset Repository},
+  howpublished = {\url{https://huggingface.co/datasets/RustBustersHSV/laser-cleaning-qa}},
+  license = {MIT}
+}
+```
+
+## Dataset Curators
+
+This dataset was created by RustBustersHSV to support the development of their customer service AI assistant.
+
+## Licensing Information
+
+This dataset is licensed under the MIT License, which allows for both commercial and non-commercial use, modification, distribution, and private use, provided that the original copyright and permission notice is included in all copies or substantial portions of the dataset.
+
+## Contributions
+
+Thanks to the team at RustBustersHSV for developing the templates and domain expertise that made this dataset possible.