• DOI:10.48550/arXiv.2208.06032
• Corpus ID: 251554795

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

@article{Singh2022CORNETAN,
  title={CORNET: A neurosymbolic approach to learning conditional table formatting rules by example},
  author={Mukul Singh and Jos{\'e} Pablo Cambronero S{\'a}nchez and Sumit Gulwani and Vu Le and Carina Negreanu and Mohammad Raza and Gust Verbruggen},
  journal={ArXiv},
  year={2022},
  volume={abs/2208.06032}
}

• Mukul Singh, José Pablo Cambronero Sánchez, +4 authors Gust Verbruggen
• Published 2022
• Computer Science
• ArXiv

Spreadsheets are widely used for table manipulation and pre- sentation. Stylistic formatting of these tables is an important property for both presentation and analysis. As a result, pop- ular spreadsheet software, such as Excel, supports automatically formatting tables based on data-dependent rules. Un- fortunately, 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 C ORNET , a neuro-symbolic system that tackles the novel problem of auto- matically learning such formatting rules from user examples of formatted cells. C ORNET 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 C ORNET to a wide range of symbolic and neural baselines. Our results show that C ORNET can learn rules more accurately, across varying conditions, compared to these baselines.