Dataturks.com

What do you like best?

We've been using dataturks for tagging training data quickly for NLP and computer vision models. The platform is great and allows us to tag larger amounts of data at much faster speeds. Our tagging speed is like 5x what we used to be at. It's also way faster to set up projects compared to systems we've used before and we've seen enormous improvements in precision/recall in our model because of the training data and feedback loop we now have.

Some of the main reasons we used dataturks after trying out their platform and several others:

-- Ease to set it up initially and for each new dataset

-- Some of features that make it really easy for taggers like keyboard shortcuts

-- API that can integrate easily into our workflow

-- Very intuitive tool - I'm not technical but I can configure each project and set up just about all the requirements on my own.

-- Took about 5 minutes to kick off - other ones we tried took far longer to set up

What do you dislike?

Not much! There may be little bells and whistles missing but they're growing really fast and have been very responsive and have quick turnaround any time there's an important feature that isn't in the platform that fits into the use cases people use dataturks for.

What problems are you solving with the product? What benefits have you realized?

Getting large amounts of data tagged quickly and correctly for training deep learning NLP and computer vision models.

Much faster time to tag data and train new models.

We were able to implement feedback loops and continuously add new data as a workflow for taggers, allowing our models to improve over time.

The bottom line benefit is that we were going to build this in house but then found dataturks and their product impressed us.