Datasaur

Manage your entire data labeling workflow with a single tool

Datasaur builds data labeling tools for natural language processing. Using intelligent models and improved workforce management, Datasaur helps ML leaders better manage their team and improve the quality of their training data.
Discussion
Would you recommend this product?
9 Reviews5.0/5
🚀After a year of incubation, very excited to finally launch Datasaur to the public! In my previous companies (Apple, Yahoo) I realized my team was spending an outsized amount of time and effort in gathering labeled data. It felt like startups and large companies alike were re-inventing the wheel when it came to building out their labeling processes. 🦕We started Datasaur to create the single most comprehensive solution for the industry. We've interviewed and spoken with 100+ AI teams globally and have encapsulated best practices. Multiple labelers can be assigned to the same project to improve accuracy and reduce human bias. Instead of labeling common terms like "New York" and "Starbucks" from scratch, intelligence can help automate a portion of the work and give labelers a head start. And everything is built from the ground up for power users, complete with keyboard shortcuts and helpful extensions. So far we've helped teams label and detect "fake news", improve customer service chatbots and better understand customer reviews. We believe Natural Language Processing (NLP) is in a great position to make changes across many industries. 💬Really looking forward to feedback from this community! How have you dealt with labeling processes in your company? What would you like to see in an ideal solution? If you have any questions about labeling or NLP in general, happy to discuss here!
Upvote (11)Share
Early beta tester here - really impressed by the product Ivan and team have built. If you're someone that loves keyboard shortcuts (I'm looking at all you Alfred or Gmail power-users) Datasaur will give you so much power underneath your fingertips. Highly recommend giving it a try!
Thanks @victor_zhang1 ! Look forward to continuing to work together.
Also an early beta tester: This is such an improvement from our previous NLP data labeling workflow, which involved both spreadsheets and a hacked-together version of Doccano. Really useful for both document-level and word-level labeling. In the near future, my team will be training custom dependency parsers and coreference resolution tools, and Datasaur has really nice features for that kind of labeling. And the keyboard shortcuts are awesome. Definitely worth trying.
@julie_mecca thank you very much 🙏look forward to what we'll be able to achieve together!
Neat tool! I’ve done some labeling on the past on some open source tools but they weren’t getting updated very frequently. Will give this a shot!
Thanks @trq__ - that's where we were at too. We also are a big fan of the existing open source solutions, and just wanted to take it to the next level commercially.