products that directly match the query. The query and the name of a relevant product may not share words. So if the product is in the same category as the query but has different specifications such as brand, it is still relevant. Example of a relevant judgment: the product 'Gifford Desk' is relevant to the query 'writing deskʼ. The user searched for a writing desk and the returned product is a desk from the brand Gifford that can be used to write. Therefore, the product matches the userʼs intent. Examples of relevance judgments: ('dining table vinyl cloth', 'horwich dining table', 'Not relevant') ('togo chair', 'evendale upholstered side chair', 'Relevant') Relevantの定義 Relevantの例 適合性判定の例 *https://github.com/danimtk/chatGPT-goes-shopping/blob/main/data/guidelines/human_baseline_ten_shot_WANDS.txt