Add Claude 3.5 Haiku synthetic text2cypher dataset

[IamAGP]

Summary

Adds a new synthetic text2cypher dataset generated using Claude 3.5 Haiku via AWS Bedrock, expanding the model coverage in this repository.

What's New

• New model: First Claude 3.5 Haiku dataset in the collection
• Budget-friendly: Configurable sample sizes with cost estimation
• Enhanced validation: Improved false_schema detection (goes beyond what opus notebook implements)
• Methodology: Follows established patterns from existing synthetic datasets

Files Added

• datasets/synthetic_haiku35_demodbs/ - Complete dataset folder
• README.md - Documentation following project conventions
• anthropic_text2cypher_haiku35.ipynb - Generation notebook with AWS Bedrock integration
• text2cypher_haiku35.csv - Sample dataset (10 questions for validation)
• Supporting CSV files (text2cypher_questions.csv, text2cypher_schemas.csv)

Testing

• ✅ Tested with 10 sample questions
• ✅ All validation metrics working (syntax_error, timeout, returns_results, false_schema)
• ✅ Successfully connects to AWS Bedrock Haiku 3.5
• ✅ Neo4j demo database validation working

Value Added

• Provides cost-effective alternative to larger models
• Enables comparison studies between different model sizes
• Maintains research reproducibility with clear methodology
• Ready for larger scale generation when needed