Hi! Wanted to give a template idea. It would be great to have a data labeling template! If I could scroll through each example and as I go be able to label it I would certainly use it, and I’m sure other data science teams would as well. Also, would be great to assign different samples to different team members so it’s easy to split the work.
I’m a data scientist at Coda and love this idea. We’ve built models here at Coda where our training sets were manually labeled by some of us. I imagine this could work in a pretty straightforward way – you could have a select list of labels that you could choose from and assign labels from that set. And there could definitely be a random (if desired) assignment of new rows to team members.
How were you thinking about building this? What sort of use case did you have in mind?
Thanks for reaching out,
Yeah that sounds good. Labels could be randomly assigned to different team members.
I would try to make the class selection very easy, like a checkbox or something. And if it works well on the phone that would be even better that way you can label when you’re on the go.
Also would add a small text field so you can write any notes about your label that could be useful.
Could even hook it up with Slack so that if you’re waiting for someone else to label, you could get notified when they finished labeling or something.
My use case would be sentence classification. I hate labeling on Google Sheets. So I’d like to scroll through sentences on my phone or some nice UI, and label them as I go. This is a bit more complex, but many times is good to check your labels against other people in your team to make sure you’re thinking of the same definition, so I’d assign some of the samples to be labeled by all team members.
By the way I did it and it works great!
Very nice to do it on the phone, easy to scroll through different data points.
That’s fantastic. I love how easy it is to scroll through the mobile pages.