I believe there is always some innate error in the relevance but in this blog post are some ways to improve the relevance by leveraging contextual information.
Hope this helps make the recommendations more accurate and here is an additional resource to help you think about how you'd define/determine relevance.
Can AWS Personalize be optimized/segmented for cachable recommendations?asked 7 months ago
Should ECS/EC2 ASGProvider Capacity Provider be able to scale-up from zero, 0->1Accepted Answerasked 9 months ago
How quickly do the events from putEvents influence recommendations ?asked a year ago
Giving weights to event types in amazon personalizeasked 10 months ago
Amazon Personalize- Exploration weight impact on recommendationsasked 16 days ago
How to avoid Personalize SIMS to recommend only popular items?Accepted AnswerEXPERTasked 2 years ago
Personalize filters - Users metadataasked 2 months ago
Can we/How to provide an array as contextual metadata on GetRecommendation?asked 7 months ago
Including negative feedback in AWS PersonalizeAccepted Answerasked a month ago
Personalize Recommendations for Anonymous Users using SessionIdasked 16 days ago