Multiple selections in smart text editing

Multiple selections, though heavily used in file managers and drawing editors, are virtually nonexistent in text editing. This paper describes how multiple selections can automate repetitive text editing. Selection guessing infers a multiple selection from positive and negative examples provided by the user. The multiple selection can then be used for inserting, deleting, copying, pasting, or other editing commands. Simultaneous editing uses two levels of inference, first inferring a group of records to be edited, then inferring multiple selections with exactly one selection in each record. Both techniques have been evaluated by user studies and shown to be fast and usable for novices. Simultaneous editing required only 1.26 examples per selection in the user study, approaching the ideal of 1-example PBD. Multiple selections bring many benefits, including better user feedback, fast, accurate inference, novel forms of intelligent assistance, and the ability to override system inferences with manual corrections.