gMeta: Template-based Regular Expression Generation over Noisy Examples

Regular expression is a technology widely used in software development for extracting textual data, validating the structure of textual documents, or formatting data. Existing regular expression generation works always assumes that the examples are faultless, however, in real industrial scenarios, this assumption does not fully hold. In this paper, we present a simple but effective templated-based approach to generate regular expressions over noisy examples. Specifically, we design an abstract data form (namely, MetaParam) to approximately describe and cluster the input examples. Then, we propose a practical dynamic thresholding scheme to filter out anomalous examples. Finally, we design a template-based regular expression generation algorithm, which is efficient, interpretable and extensible. We performed an experimental evaluation on two different extraction tasks applied to realworld datasets and obtained promising results in terms of precision. Moreover, gMeta achieves excellent results in real industrial scenarios.