Adding structured and unstructured metadata to your iteractions, items, and users datasets can improve the relevance of recommendations provided by Amazon Personalize. Selecting the right metadata and properly preparing it for Personalize is an important task.
The notebooks below demonstrate features specifically designed for working with metadata fields.
The item-text-unstructured-metadata.ipynb provides an example of how to include unstructured text as item metadata and the impact it can have on the relevance of recommendations.
This sample code is made available under a modified MIT license. See the LICENSE file.