FABRIC: Fully-automated broad intent categorization in e-commerce

Predicting the user’s shopping intent is a crucial task in e-commerce. In particular, determining the product category, which the user wants to shop, is essential for delivering relevant search results and website navigation options.

Existing query classification models are reported to have excellent predictive performance on the single-intent queries (e.g. ‘running shoes’), but there is little research on predicting multiple-intents for broad queries (e.g. ‘running gear’). While training data for broad query classification can be easily obtained, evaluation of multi-label categorization remains challenging, as the set of true labels for multiintent queries is subjective and ambiguous.

In this work we propose FABRIC – an automatic method of creating evaluation data for multi-label e-commerce query classification. We reduce the ambiguity of the annotations by blending the label assessment from three different sources: aggregated click data, query-item relevance predictions and LLM judgments