Keyphrase Generation

Keyphrase generation (KPG) aims to automatically create a set of phrases summarizing a document's core concepts, a crucial task for information retrieval and text understanding. Current research emphasizes improving both the recall and precision of generated keyphrases, often employing large language models (LLMs) within generate-then-select frameworks or through contrastive learning methods to enhance phrase-level representation. These advancements, along with explorations into unsupervised domain adaptation and handling long documents, are improving the accuracy and applicability of KPG across diverse domains, including scientific literature and legal texts.