We provide annotators with annotation guidelines, as illustrated in Figure(/Annotation.png). All image-text pairs have undergone pre-screening to confirm their metaphorical nature. Annotators are tasked with extracting items from the source and target domains based on the provided image-text pairs, using only a single Chinese vocabulary term to describe each domain. Typically, the annotators determine the target and source domains by observing two conflicting objects.
	• Target domain identification: Most instances can be found in the image or text, typically related to the objects truly described in advertisements. For example, the target domain in commercial advertisements is usually related to the products.
	• Source domain identification: The source domain is highlighted by the shape or related attributes that conflict with the target domain. And most instances also appear in the image or text.
	Additional examples are provided in Figure(/exp1.png, /exp2.png, /exp3.png, /exp4.png) for reference.