Generating Scientific Definitions with Controllable Complexity

• DOI: 10.18653/v1/2022.acl-long.569
• Publication Date: 2022-06-03T01:34:53Z
• Authors: Tal August; Katharina Reinecke; Noah Smith
• Keywords: Sequence (biology)

Unfamiliar terminology and complex language can present barriers to understanding science. Natural processing stands help address these issues by automatically defining unfamiliar terms. We introduce a new task dataset for scientific terms controlling the complexity of generated definitions as way adapting specific reader's background knowledge. test four definition generation methods this task, finding that sequence-to-sequence approach is most successful. then explore version in which are at target level. novel reranking find human evaluations it offers superior fluency while also complexity, compared several controllable baselines.