If you've ever tried to build an email parser, fine-tune an LLM on corporate communications, or benchmark a RAG (Retrieval-Augmented Generation) pipeline, you've likely run into the infamous Enron dataset.
For over two decades, Enron has been the undisputed "gold standard" for open communication data. But let's be honest with ourselves: it’s outdated, legally complicated, and filled with massive privacy risks (PII).
We faced this exact bottleneck while stress-testing our own data pipelines and AI systems. Real corporate data is locked behind strict NDAs and privacy laws, and Enron just wasn't cutting it anymore for modern tech stacks.
So, we decided to fix it. Today, we are releasing a completely free, open-source, and high-quality synthetic dataset containing 1.25 million modern communication records.
📊 What's Inside the Dataset?
Unlike older datasets that only focus on raw email text, we wanted to mirror how modern teams actually communicate. The 1.25M records are split across three major pillars:
- Emails
- Chats
- Calendar Events
Best of all: It is 100% free of real-world PII (Personally Identifiable Information), making it fully compliant and safe to use in any environment.
🛠️ Get Started Right Now
The dataset is fully open, hosted on HuggingFace, and ready for you to dive in. The documentation and schema details are already live in our repository's README.
👉 https://huggingface.co/datasets/sinabis-group/sinabis-synthetic-corporate-dataset
We would love to hear from the community! How are you currently solving the lack of open text datasets for communication?
Drop your thoughts, feedback, or questions in the comments below!
Top comments (0)