Rethinking the Objectives of Extractive Question Answering

TL;DR

This paper challenges the independence assumption in extractive question answering models and introduces a joint probability modeling approach with a compound objective, improving accuracy across multiple models and datasets.

Contribution

It proposes a novel joint probability modeling method with a compound objective, surpassing traditional independence assumptions in extractive QA.

Findings

Compound objective improves exact match scores.

Independence assumption causes common errors.

Method effective across multiple models and datasets.

Abstract

This work demonstrates that using the objective with independence assumption for modelling the span probability $P(a_s,a_e) = P(a_s)P(a_e)$ of span starting at position $a_s$ and ending at position $a_e$ has adverse effects. Therefore we propose multiple approaches to modelling joint probability $P(a_s,a_e)$ directly. Among those, we propose a compound objective, composed from the joint probability while still keeping the objective with independence assumption as an auxiliary objective. We find that the compound objective is consistently superior or equal to other assumptions in exact match. Additionally, we identified common errors caused by the assumption of independence and manually checked the counterpart predictions, demonstrating the impact of the compound objective on the real examples. Our findings are supported via experiments with three extractive QA models (BIDAF, BERT, ALBERT) over six datasets and our code, individual results and manual analysis are available online.