READR: A RST Driven Text Processing Platform for Reading Comprehension
CitationCui, Jake Yee. 2020. READR: A RST Driven Text Processing Platform for Reading Comprehension. Bachelor's thesis, Harvard College.
AbstractHuman language is built upon underlying structure, yet the text we are used to reading is displayed in a flat body with minimal context. I present READR, a tool that can identify and highlight hidden structures in text (like examples and elaboration) to augment reading comprehension in realtime. READR classifies sentences based on the linguistic Rhetorical Structure Theory (RST) by deep learning a neural embedding across >10000 labeled corpora. Given any target text document, whether it be a research paper, Wikipedia article, or textbook, READR presents an interactive interface in which users can quickly skim and process text based on rhetorical function. I show that READR has the potential to improve comprehension and reduce reading time in a preliminary user study with five participants.
Citable link to this pagehttps://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37364667
- FAS Theses and Dissertations