“…These approaches are not directly applicable to our problem setting since we require short phrases focusing on the performance of an action by the user. Approaches such as TextRank (Kazemi et al, 2020), SingleRank (Wan and Xiao, 2008), Ex-pandRank (Wan and Xiao, 2008), TopicRank (Bougouin et al, 2013), TopicalPageRank (Sterckx et al, 2015), PositionRank (Florescu and Caragea, 2017), Bi-LSTM-CRF Sequence Labeling (Alzaidy et al, 2019), FACE (Chau et al, 2020), and Mul-tipartiteRank (Boudin, 2018) have been applied to the task of key phrase extraction from documents as opposed to theme detection from clusters. For cluster-labeling, there has been work around using keyword extraction from clusters, utilizing WordNet synsets to expand the keywords, followed by a selection procedure to assign the final label (Poostchi and Piccardi, 2018;Chang and McKeown, 2019).…”