In this paper. we explore the approaches to the problem of cross-domain few-shot classification of sentiment aspects. By cross-domain few-shot. we mean a setting where the model is trained on large data in one domain (for example. https://balboasurfandsstylers.shop/product-category/dress/