From Documents to Structured Data with Interactive Labelling
DeepTagger is a no-code platform that makes your judgment scalable. It uses your annotations as an example to extract information from new documents. Highlight what matters to you once, and let DeepTagger handle the rest with precision. API access included.
While analyzing the Enron Email dataset for a PhD project, we needed to extract data from hundreds of thousands of emails in various formats, and then trace chains that included incidents of “knowledge hiding.” But we got stuck on the very first task: splitting long email chains into individual emails.
Custom Python parsers failed.
RegEx broke.
Traditional ML tools, such as spaCy Prodigy, or Label Studio, couldn’t handle the complexity 🤯 Doing it manually would have meant admitting defeat.
So we built our own annotation tool that could handle nested data structures 🛠️. However, even with perfect annotations, traditional models couldn’t generalize — the data was too diverse, and the examples were too few.
Then OpenAI posted "Introducing Structured Outputs in the API," and everything clicked ⚡ Our annotations became few-shot examples instead of training data. ✅ No model training needed — just smart prompting.
That’s when we realized this could compete with traditional OCR tools by offering a completely different experience.
A few months of polish later… Deeptagger was born 🚀 Hope you love it! ❤️
This product was born out of real-life problems 📨
While analyzing the Enron Email dataset for a PhD project, we needed to extract data from hundreds of thousands of emails in various formats, and then trace chains that included incidents of “knowledge hiding.”
But we got stuck on the very first task: splitting long email chains into individual emails.
Custom Python parsers failed.
RegEx broke.
Traditional ML tools, such as spaCy Prodigy, or Label Studio, couldn’t handle the complexity 🤯
Doing it manually would have meant admitting defeat.
So we built our own annotation tool that could handle nested data structures 🛠️. However, even with perfect annotations, traditional models couldn’t generalize — the data was too diverse, and the examples were too few.
Then OpenAI posted "Introducing Structured Outputs in the API," and everything clicked ⚡
Our annotations became few-shot examples instead of training data.
✅ No model training needed — just smart prompting.
That’s when we realized this could compete with traditional OCR tools by offering a completely different experience.
A few months of polish later… Deeptagger was born 🚀
Hope you love it! ❤️