The Secret Recipe for Data Annotation At Scale: The Odetta Difference

October 29, 2020 Tayyaba Qamar

When Odetta was first getting started, I remember one of my data scientists friends, John, suggested we become a full-stack data science company, and start with data annotation. We had done some annotations, but it seemed like a competitive marketplace. Data annotation is a way of processing text, images, videos so it gets classified or marked, and can become part of a team's artificial intelligence (AI) algorithm for machine learning efforts. For example, we have tagged electric bills manually by identifying the sum totals so machines can be trained to do this work by our clients. We have also summarized text and score summaries prepared by machines so that machines can learn to read and write quickly. Although I was initially worried about the complexity of the AI marketplace, what we have learned is that there was a real need for trained annotation teams that can deliver an Account-Managed Service at the fast pace that AI-driven, venture-backed companies require.

Over the past two years, we have been perfecting the recipe for data annotation as a service and are really excited about the results we are delivering for clients. A good example is our work with Primer.ai, a machine intelligence company that builds machines that can read and write, automating the analysis of large datasets. Primer has spent the past two years working with a team of linguists on Upwork and before that, they had worked with Figure8, but found it very pricey. Their Upworkians had great performance, but it was hard to scale and required the time of one of their more valuable data science resources to manage the Upwork team basically full time. That's an expensive role to pay for. About one year ago, Primer.ai turned to Odetta, and we took over the account management and built a team of 12 (now 20) linguists who serve Primer.ai on a daily basis.

According to Charlene Chambliss, Data Scientist at Primer.Ai,

Any Data Scientist /Machine Language professional knows how much of a headache it can be to get high-quality labeled data. There are plenty of offerings in this space, and yet none of them were quite the right fit for us. On the one hand, you can spend $10k with one of the large providers to get 100,000 labels for a single project (a volume which we would never realistically need). Or on the other hand, you can bring everything in-house and have more flexibility, but now it's someone's full-time job to manage data labelers, and that has its own costs. And, forget about trying to farm tasks out to Amazon Mechanical Turk (MTurk.com), if you're solving complex problems that require any amount of care and attention. Odetta met us right at that Goldilocks zone. With Odetta, we have tight communication loops and can iterate quickly, while getting our data labeled for a wide variety of challenging tasks to meet our deadlines. And importantly, we know our annotators are paid fairly for their work

But how exactly does Odetta manage to outperform on data annotation services? Our secret recipe is comprised of our:

Powerful Odetta SaaS outsourcing platform that removes all the inefficiencies that have plagued traditional outsourcing and freelance platforms.
Unique Odetta client project management process which is highly automated and includes built in quality assurance mechanisms, and
Most importantly, our superstar Odetta team (Odettians) of highly skilled women in developing countries across the globe.

Odetta has been delivering results for large scale data science and annotation projects since its inception over two years ago. Our approach and methodology for managing annotation and Natural Language Processing (NLP) projects is also quite unique and has been time tested and proven. Every project is tailored to client needs and follows our client kickoff process and toolkit to ensure rapid results at scale:

To find out more about Odetta comprehensive solutions can help your business with it’s data, sales, recruiting and operation needs, just contact us via this form and we’ll reach out to discuss your needs.