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.
Just tested this on a messy contract doc, blown away by how well it picked up the structure after just a couple of highlights. Way better than anything I’ve tried with Prodigy or Label Studio.
Wow, excellent! I searched some instruments for analyze a non-structure data. Will try, thx!
DeepTagger
@ivan_lindberg that’s awesome to hear 🙌 Non-structured data is exactly what we wanted to make easier to handle. Can’t wait to hear how it works out for you. Thanks for giving it a try! 🚀
DeepTagger
@ivan_lindberg Would be happy to assist you! If you have any questions or looking for a quick demo! Otherwise give it a try - it's fully self-service.
Flipner AI
Great product! Wish you TOP position!
DeepTagger
@alex_egorov thank you very much. We’ll do our best 🚀
DeepTagger
@alex_egorov Thank you, Alex! We'll keep trying!
Arbonum
DeepTagger
@arbonum really appreciate it 😊 Thanks for the support of our launch 🙌
DeepTagger
@arbonum Thank you! We really tried to make it unique!
Fluensa
Great product, love this.
DeepTagger
@goldentak Thank you for checking it out!
DeepTagger
@goldentak thank you so much, Nurasyl! 🙌 Really glad you love it, means a lot to us!
DiffSense
Besides CVs. What are the the top 3 other document types people use this for?
DeepTagger
@sentry_co
1) Job Postings. In this case we're extracting employer details, which is objective, and skills, which is subjective, skills can mean different things to different people. DeepTagger allowed to define via examples exactly what sort of skills needed to be extracted.
2) Court Decisions. Those are lengthy and verbose, and expert lawyer annotated several documents, highlighting key parts, and then DeepTagger was able to extrapolate that, to bulk-extract that sort of information from other documents. This theoretically would've been possible to do by tweaking the prompt, but having ability to simply select key paragraphs made this much easier.
3) Contracts. Just like with Job Postings we have both objective information, like name and contacts of the parties, as well as subjective information, like "responsibility". It was much easier to define latter via examples, rather than via prompt-tweaking.
bonus ) We've used DeepTagger to train it to rate amateur prose. It took an expert about 4 hours to hand-annotate about 40 amateur poems, before DeepTagger became more-or-less reliable, since quality of a poem is extremely subjective. But eventually it worked and it was able to effectively analyse actual poems sent to a poetry competition, and its extractions (when weighted and converted to a score) highly correlated with ranking of the entrants as judged by a human judge.
DiffSense
@avloss What are some areas you have on the roadmap to tackle next. What cant the system do well today, that you want to do in the future when it matures?
DeepTagger
@sentry_co MCP integration feels like a "must have", we were even considering not launching until it's ready, but decided to go ahead. The system does what it does reasonably well, as good and as fast as LLM can extrapolate from examples. But that API integration just doesn't quite "cut" it in 2025. It has to be MCP to be mass-appealing (for non-technical business users, for instance). Plus, it should be completely transparent for an end user, perhaps instead of logging into DeepTagger and working there, user will be telling their LLM client - "I want to extract important paragraphs from documents stored in this folder", and then LLM (with DeepTagger MCP connected), will be replying "here's a suggested extraction patter" showing user a link - user would be able to view and amend that data.
DiffSense
@avloss Yepp. MCP is good path. And a lot of fun to work on. Its really about responses, error recovery, and staying within the context window before the LLM ebbs out. In your case, I think it would be massive if you add local file support for the MCP. Not that hard to add. 1day maybe. But requires the MCP to be executable binary. It can still be installed with a one liner without deps, and get updates etc. But a minimum binary is required to do the local file saving / opening. The secret sauce can still be chucked into a cloud MCP / API.
Jinna.ai
Nice. If not API, what other interfaces does your project support?
DeepTagger
@nikitaeverywhere We're currently working on MCP integration - basic prototype is already working, but not published, also we're considering integrations with enterprise business platforms and perhaps something like Zapier would also make sense.