STaRK-Amazon
STaRK-Amazon focuses on product recommendation. The dataset emphasizes customer-centric factors like product quality, functionality, and style and incorporates single-hop relational data involving brands, categories, and products in complementary or substitute relationships. Queries are presented in a conversational style, making the data more relevant to practical scenarios.
Semi-structured Knowledge Base
The Amazon SKB includes four entity types (product, brand, color, and category) and five
relationship types (also_bought and also_viewed between products, and has_brand,
has_color, and has_category linked to products).
The textual information for products is comprehensive, featuring descriptions, customer reviews, and Q&A, while other entities have their names or titles as textual attributes.
Data Statistics
| Num of entity types | 4 |
|---|---|
| Num of relation types | 5 |
| Num of avg. degree | 18.2 |
| Num of entities | 1,035,542 |
| Num of relations | 9,443,802 |
| Num of tokens | 592,067,571 |
STaRK QA Dataset
Data Statistics
| Synthesized | Human-generated | |
|---|---|---|
| Num of queries | 9,100 | 81 |
| Num of queries with multiple answers | 7,082 | 64 |
| Average num of answers | 17.99 | 19.50 |
License: CC-BY-4.0
QA Examples
Synthesized Set
Q: Can you suggest some imported women's yoga pants with a stylish screen print on the left leg?
A: 601168, 947318
Q: Could you help me find shooting sticks that have an aluminum alloy tipped foot designed for all kinds of uneven terrain?
A: 109824, 177809, 501268, 949143, 390168, 443297…
Q: Looking for Simms fishing shoes that are completely waterproof and suitable for party wear. They must have a bigger toe box, as I had issues with tight toe areas in the past.
A: 218306
Human-generated Set
Q: Can you recommend a mouse repellent that has a nice smell that can also be used as stocking stuffers?
A: 739490, 926663, 779709, 536847
Q: I've bought some women's weightlifting gloves for myself from Harbringer, and I love them. Can you find me similar gloves for my husband?
A: 6369, 103525, 418983, 910535, 119920…
Q: I like running, swimming, and biking. I love my coolmax socks from RunBreeze. Recommend me a tracksuit to wear for my upcoming competition.
A: 776529, 776514, 776518, 776511
Reference
[1] Julian J. McAuley, Christopher Targett, Qinfeng Shi, and Anton van den Hengel. 2015. ImageBased Recommendations on Styles and Substitutes. In SIGIR. ACM.