Auto-Type: Synthesizing Type-Detection Logic for Rich Semantic Data Types using Open-source Code

Given a table of data, existing systems can often detect basic atomic types (e.g., strings vs. numbers) for each column. A new generation of data-analytics and data-preparation systems are starting to automatically recognize rich semantic types such as date-time, email address, etc., for such metadata can bring an array of benefits including better table understanding, improved search relevance, precise data validation, and semantic data transformation. However, existing approaches only detect a limited number of types using regular-expression-like patterns, which are often inaccurate, and cannot handle rich semantic types such as credit card and ISBN numbers that encode semantic validations (e.g., checksum).

We developed AutoType, a system that can synthesize type-detection logic for rich data types, by leveraging code from open-source repositories like GitHub. Users only need to provide a set of positive examples for a target data type and a search keyword, our system will automatically identify relevant code, and synthesize type-detection functions using execution traces. We compiled a benchmark with 112 semantic types, out of which the proposed system can synthesize code to detect 84 such types at a high precision. Applying the synthesized type-detection logic on web table columns have also resulted in a significant increase in data types discovered compared to alternative approaches.

Our benchmarking dataset has been made available on GitHub https://github.com/congy/AutoType/ (opens in new tab) to facilitate future research.