PBS and Amazon Web Services have announced the network is using generative AI to provide enhanced search results to viewers on the PBS App and PBS LearningMedia platforms.
To improve its search capabilities, PBS said it teamed up with AWS science and strategy experts from the AWS Generative AI Innovation Center. They developed and launched its new generative AI-powered search engine powered by the AWS managed generative AI service Amazon Bedrock. Working backwards from their problem with the AWS team, PBS envisioned the updated search engine, trained and tested models, then cataloged and tagged metadata for more than 700,000 assets in less than six months. As a result, PBS viewers can now search the entire PBS library based on their interests—across subgenres, topics, themes, and even moods.
“If we did manual metadata-tagging for this project, it would have taken years,” theorized Mikey Centrella, director of product at PBS, Digital Innovation. “With Amazon Bedrock, we were able to do it quickly, at a fraction of the cost and within the secure AWS environment we already have. In just a few months, we built a proof-of-concept, tested and successfully scaled it into production. By greatly improving our search tools, we’re delivering a better experience for viewers and showing them different parts of our catalog that they might not have found otherwise.”
The deployment and the technology used for the improve search systems was laid out by AWS in a blog, by Zach Dugan is a senior industry product marketing manager for Media & Entertainment with AWS.
The deployment builds on a relationship between PBS and AWS that goes back more than a decade. PBS already uses AWS across nearly every aspect of its operations, which enables it to cost-efficiently scale and deploy new technology.
“One of the great things about building on AWS is that we can easily explore ways to modernize our operations, without impacting what’s already in production,” said Centrella. “Our assets are already in the cloud, so it’s a safe way to play, figure out what we might want to invest in and then apply it across the business. As a nonprofit, we must be smart about our choices and use our dollars efficiently. AWS helps us protect the value of what we’re creating for the American public.”
Dugan explained that Centrella’s team wanted to create a solution with generative AI that would add value and have an immediate impact for viewers. After connecting with the AWS Generative AI Innovation Center in late 2024, the team began to explore how best to augment the existing search engine of PBS—built on AWS and cloud-native tools from PBS—using generative AI.
“The table stakes [for streaming services] have changed because of technology,” noted Centrella. “Our users wanted a more robust search engine, so that’s what we created.”
After defining the project scope, Dugan reported that PBS began an eight-week engagement with the AWS Generative AI Innovation Center.
The mission of the Generative AI Innovation Center is to help organizations, like PBS, accelerate their adoption of generative AI to unlock mission-changing outcomes. They accomplish this through bespoke generative AI proof-of-concepts, designed in collaboration with customers, according to AWS.
Using their expertise at AWS, Centrella’s team began building the new search engine framework and testing different models. The existing search engine implementation of PBS had basic logic functionality but the use of generative AI for extensive metadata tagging enabled the new search engine to offer much more specific results.
For example, each area of search classification now encompasses hundreds of options and dozens of metadata tags that can be applied to any given project. For example, a viewer can search heartwarming in the PBS App and receive a result such as All Creatures Great and Small, a charming series about a veterinarian in 1930s Northern England, AWS explained.
“Generative AI is great at pattern recognition and summarization, and metadata—set up correctly—provides a solid foundation for success,” Centrella explained. “We found that the show descriptions and transcripts gave us the most value for the price for analyzing content. We narrowed down four areas of classification internally, then used models on Amazon Bedrock to analyze the content and apply the most appropriate metadata.”
In addition, AWS and PBS established a sandbox environment to collaboratively evaluate model performance using Amazon Bedrock, ultimately determining the Anthropic Claude Sonnet to be the best fit.
"The PBS team then ran parallel workstreams on 50 shows, with PBS colleagues manually creating metadata tags based on show scripts to compare with the results created using generative AI," the blog noted. "Looking at the outcomes, they found that the model produced data sets as good as or better than the ones that were created manually, and doing so in minutes. Based on those promising results, PBS moved the model into operation, generating metadata tags for its entire content library of 700,000 assets and incorporating this new metadata into its search engine."
Looking forward Dugan noted in the blog that “the PBS Innovation Team is working again with the AWS Generative AI Innovation Center to evaluate other ways to incorporate generative AI into its infrastructure. The most likely next step is to extend enhanced search recommendations to the network’s kid’s catalog.”
“PBS understands that the audience and their behaviors are changing, across all the available streaming options, gaming platforms and mobile devices. We live in a multiplatform, multimedia world, and we want to meet people where they are,” shared Centrella. “It can be challenging to address all the new technology and distribution types that seem to emerge daily, and at the same time, linear still matters. AWS makes it easy for us to evolve and experiment, while maintaining our core values.”
More information about building scalable generative AI applications on AWS is available here or here.
