Amazon has created a dataset of difficult search queries from online shoppers to help train product search models and advance research in ranking strategies. The dataset contains 97,345 queries in English, 15,180 queries in Spanish, 18,127 queries in Japanese, a list of up to 40 potential results for each query, and a judgment labeling the relevance of results as exact, substitute, complement, or irrelevant to the query.
Image credit: Flickr user Robbert Noordzij