Skip to content

Suggest labeling functions #6

Open
@evidencebp

Description

@evidencebp

A labeling function is a computable way to automatically label certain cases.
These labeling function are not of 100% accuracy, but they don't need to be in order to be useful.
Agreements and disagreements between labeling function can be used for active learning, and find efficiently informative functions.

A labeling function that agrees with the concept can be used to find false negatives.
For example, a commit introducing a file is more likely to be adaptive.
Hence, the list of file first commits that are not already labeled as adaptive are likely to provide false negatives and improve recall.

A labeling function disagreeing with the concept can be used to find false positive.
For example, a commit labeled as adaptive that modifies only non source code files, is more likely to be a false positive.
Such examples are helpful in improving the precision.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions