The AnyClip Recommendation Engine seamlessly delivers video content tailored to the webpage context. By leveraging advanced machine learning and a robust video library, it ensures the site audience experiences the most relevant and engaging videos. Here's how it works.
Step 1: Analyze the Page
AnyClip Recommendation Engine begins by scanning the webpage to understand its content. It collects essential elements like:
Page metadata.
Page sections and their content.
This foundational step builds a comprehensive view of the page.
See the list of parsed HTML metadata attributes in the section HTML metadata attributes.
Step 2: Build Vector Embeddings
AnyClip Recommendation Engine builds the page vector embeddings based on the metadata and textual data from Step 1.
Note
AnyClip Recommendation Engine builds video vector embeddings for all videos when analyzing them, so at the moment of video search, the videos' vector embeddings are already calculated.
Step 3: Match Videos to the Page
AnyClip Recommendation Engine compares vector embeddings of the page and stored videos' vector embeddings, and returns the most relevant videos.
In the end, it filters the videos according to the page's player configuration filters.
HTML metadata attributes
The following page metadata attributes are parsed by on Analyze Page step:
og:title - The page title
og:description - The page description
og:site_name - The site name
keywords - The page keywords
Note
AnyClip Recommendation Engine allows configuring per site, so the page attributes above could be taken from other properties.
Contact your Customer success engineer to configure it properly.