Zero-Shot Multi-Label Topic Inference with Sentence Encoders

In this paper, we focus on Zero-shot approaches 2018b)] for topic inference tasks and subsequently, (Yin et al., 2019; Xie et al., 2016; Veeranna establish a benchmark for future study in this crucial et al., 2016) for inferring topics from documents direction. To achieve this, we conducted extensive where document and topics were never seen experiments with multiple real-world datasets, previously by a model. Furthermore, for developing including online product reviews, news articles, Zero-shot methods, we exclusively focus on and health-related blog articles. We also implemented leveraging the recent powerful sentence encoders.