Reproduce “Generation and annotation of item usage scenarios in e-commerce using large language models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Madoka Hagiri, Kazushi Okamoto, Koki Karube, Kei Harada, Atsushi Shibata
The University of Electro-Communications research explores generating item usage scenarios for e-commerce complementary recommendations using a Large Language Model (LLM). The study demonstrates that GPT-4o-mini can produce plausible usage scenarios for product categories, with 82.9% of generated scenarios rated as plausible by human annotators, laying groundwork for more context-aware and explainable recommendation systems.
Key insights
- Large Language Models can effectively generate realistic and plausible item usage scenarios solely from product category information, simulating human cognitive processes for identifying item complementarity.
- Human judgments of scenario plausibility exhibit significant subjectivity and individual variability, highlighting the nuanced nature of complementary relationships.
- A simple filtering strategy, selecting only the first few generated scenarios, can substantially improve the overall quality and reliability of LLM-generated usage data by mitigating positional bias.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.