Described are systems and methods for product recommendation refinement in a topic-based virtual storefront embedded in a topical community web page. A system and method may facilitate determining user and community member activity in the virtual storefront based on which weighted keywords are derived. A topic set containing various weighted keywords may be iteratively configured for extracting and ordering one or more products that are extracted from a plurality of heterogeneous sources. A server may be configured for refining the topic set by modifying a baseline session keyword weight or a baseline contextual keyword weight based on a total topic set weight, an elasticity parameter, core topic keyword weights associated with core topic keywords, session keyword weights, and/or contextual keyword weights; and identifying, based at least on the configured topic set, products to be presented to a user as product recommendations in the virtual storefront.