CORNET: A neurosymbolic approach to learning conditional table formatting rules by example

Spreadsheets are widely used for table manipulation and presentation. Stylistic formatting of these tables is an important property for both presentation and analysis. As a result, popular spreadsheet software, such as Excel, supports automatically formatting tables based on data-dependent rules. Unfortunately, writing these formatting rules can be challenging for users as that requires knowledge of the underlying rule language and data logic. In this paper, we present CORNET, a neuro-symbolic system that tackles the novel problem of automatically learning such formatting rules from user examples of formatted cells. CORNET takes inspiration from inductive program synthesis and combines symbolic rule enumeration, based on semi-supervised clustering and iterative decision tree learning, with a neural ranker to produce conditional formatting rules. To motivate and evaluate our approach, we extracted tables with formatting rules from a corpus of over 40K real spreadsheets. Using this data, we compared CORNET to a wide range of symbolic and neural baselines. Our results show that CORNET can learn rules more accurately, across varying conditions, compared to these baselines. Beyond learning rules from user examples, we present two case studies to motivate additional uses for CORNET: simplifying user conditional formatting rules and recovering rules even when the user may have manually formatted their data.