| Current intelligent writing assistance tools (e.g. Grammarly, Turnitin, etc.) typically work by locating the problems of essays for users (grammar, spelling, argument, etc.) and providing possible solutions. These tools focus on providing feedback on a single draft, while ignoring feedback on an author's changes between drafts (revision). This thesis argues that it is also important to provide feedback on authors' revision, as such information can not only improve the quality of the writing but also improve the rewriting skill of the authors. Thus, it is desirable to build an intelligent assistant that focuses on providing feedback to revisions.;This thesis presents work from two perspectives towards the building of such an assistant: 1) a study of the revision's impact on writings, which includes the development of a sentencelevel revision schema, the annotation of corpora based on the schema and data analysis on the created corpora; a prototype revision assistant was built to provide revision feedback based on the schema and a user study was conducted to investigate whether the assistant could in uence the users' rewriting behaviors. 2) the development of algorithms for automatic revision identification, which includes the automatic extraction of the revised content and the automatic classification of revision types; we first investigated the two problems separately in a pipeline manner and then explored a joint approach that solves the two problems at the same time. |
