Case Study: GumGum achieves 10x faster model development and high-quality multilingual data annotation with Appen

GumGum selected Appen for its robust training data platform and the Machine Learning (ML) Assisted Data Annotation

GumGum, an AI company specializing in computer vision and natural language processing for contextual ad placement, needed large volumes of accurately annotated text and images to train its models. With only two full-time annotators able to deliver about 15,000 text rows or 50,000 images per month, GumGum faced a bottleneck that slowed model development and limited its ability to detect unsafe content, classify faces and objects, and support multilingual projects. To solve this, GumGum selected Appen and its training data platform, including Machine Learning (ML)-Assisted Data Annotation.

Appen implemented a scalable annotation solution with an easy-to-use platform, ML-assisted workflows, and native-language annotators for Spanish, French, German and Japanese. As a result, GumGum can now label roughly 10,000 rows in days or hours instead of weeks, has sped model development about 10x, and freed data scientists to focus on research rather than manual labeling; Appen’s responsive support and customizable job design also improved annotation quality and throughput for GumGum’s CV and NLP projects.

GumGum

Erica Nishimura

Data Curator