Overview
A career spent working with photos, videos, music, and books has provided me ample opportunities to experiment with AI solutions for content tagging and metadata generation. From counting the number of people in a photo to applying keywords to kittens to writing fun facts about pop stars, AI-powered tools can perform a lot of mundane tasks. I’ve evaluated the work of these “robots” for over a decade, and I’ve recommended ways to leverage automation for scale, while still adhering to the level of quality demanded by human end users.
The Situation
Scholastic offers a wide range of digital products for education that will be featured on a unified new platform. There will be many types of content represented, from magazine articles to quizzes to lesson plans to slideshows to books. All content will need to be consistently tagged with metadata including topics, skills, and educational standards to power end user discovery.
The Solutions
A knowledge graph with a customized and evolving semantic layer will provide the foundation for the platform. To populate the graph, AI technologies are inserted at multiple ingestion points to speed up the process of identifying topics, aligning content to educational standards, and ultimately, tagging content with topics, skills, and standards for retrieval.
AI is helping create a robust taxonomy for topics and skills and to identify potential relationships to formalize ontological properties, based on an analysis of the content itself and the connections uncovered in the process. AI is helping to recommend educational standards that align to content, so both strong matches and the more nuanced correlations can be revealed in search.
By inserting humans-in-the-loop – editors and other subject matter experts – at the correct points in the flow, the AI-suggestions can be vetted for accuracy and completeness so that users can explore the platform knowledge graph for serendipitous discovery.