Paper: Robust Logistic Regression using Shift Parameters

ACL ID P14-2021
Title Robust Logistic Regression using Shift Parameters
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 2014

Annotation errors can significantly hurt classifier performance, yet datasets are only growing noisier with the increased use of Amazon Mechanical Turk and tech- niques like distant supervision that auto- matically generate labels. In this paper, we present a robust extension of logistic regression that incorporates the possibil- ity of mislabelling directly into the objec- tive. This model can be trained through nearly the same means as logistic regres- sion, and retains its efficiency on high- dimensional datasets. We conduct exper- iments on named entity recognition data and find that our approach can provide a significant improvement over the standard model when annotation errors are present.