In the evolving landscape of food e-commerce live streaming, the profusion of textual data, marked by an excess of promotional vernacular and unstructured formats, presents a formidable challenge for event extraction. Addressing these hurdles, we introduce a tailored ontology-based method alongside FMLEE (Food Marketing Live Event Extraction), a joint event extraction algorithm. This approach simplifies the event identification process through meticulous segmentation and the development of an ontology comprising 5 event categories and 19 argument roles. By integrating context-aware embeddings derived from pre-trained language models and applying an adversarial learning tactic, our methodology not only bolsters the robustness of our model but also significantly refines its accuracy in discerning relevant events within the scarce-resource milieu of food live streaming promotions. The effectiveness of the FMLEE model is validated by its achievement of an F1 score of 73.05%, with the inclusion of adversarial learning contributing to a 2.61% enhancement in performance. This evidences our novel contribution to the domain, offering robust technical support for the optimal exploitation of information within the sphere of food live streaming promotions. Simultaneously, this aids in the investigation of innovative applications for consumer engagement within marketing strategies and the smart regulation of marketing activities.