HumanAL: Calibrating Human Matching Beyond a Single Task

TL;DR

This paper introduces HumanAL, a method that models human annotator behavior to calibrate and improve label quality across various matching tasks using machine learning.

Contribution

It presents a novel approach to account for human error by building behavioral profiles and calibrating human input for better labeling accuracy.

Findings

Improves label quality in schema, entity, and text matching tasks

Effective across multiple domains and tasks

Utilizes black-box machine learning for calibration

Abstract

This work offers a novel view on the use of human input as labels, acknowledging that humans may err. We build a behavioral profile for human annotators which is used as a feature representation of the provided input. We show that by utilizing black-box machine learning, we can take into account human behavior and calibrate their input to improve the labeling quality. To support our claims and provide a proof-of-concept, we experiment with three different matching tasks, namely, schema matching, entity matching and text matching. Our empirical evaluation suggests that the method can improve the quality of gathered labels in multiple settings including cross-domain (across different matching tasks).