Label the data by some heuristic functions (like regular expression).
This feature is inspired by snorkel.

1. Use various simple functions to label data
2. Merge output of 1 and decide the label
3. Train the model by 2

It would be difficult to implement weak supervision on the doccano. Because weak supervision requires some programming and to execute user-defined-scripts on the application (like Jupyter Notebook) is difficult (and cause security vulnerability).

But labeling function will be useful for users to label data. To implement filter rule is one alternative.

Examples:

• Google Search Operators: The Complete List (42 Advanced Operators)
• How to use advanced search