Neurosymbolic Search Model Outperforms E-commerce Giants on Complex Queries

A San Francisco-based search and discovery company, Onton, has released a neurosymbolic model for complex product searches called Ontology 1. This model outperformed the world’s best e-commerce search engines in a benchmark test, achieving a mean precision@10 of 0.630 on a set of 90 queries. In comparison, Google Shopping and Amazon scored 0.543 and 0.469 respectively, despite indexing roughly 100% of their catalogs compared to Ontology 1’s 1%. This significant improvement in accuracy is particularly notable for complex, conversational, multimodal product searches that require a deep understanding of the query intent.