Is the Best Better? Bayesian Statistical Model Comparison for Natural Language Processing
This addresses the issue of unreliable model evaluation for NLP researchers, though it is incremental as it builds on existing statistical methods.
The authors tackled the problem of unreliable model comparisons in NLP by proposing a Bayesian statistical technique using k-fold cross-validation across multiple datasets to estimate the likelihood of one model outperforming another or achieving practical equivalence, and applied it to rank six English part-of-speech taggers across two datasets and three metrics.
Recent work raises concerns about the use of standard splits to compare natural language processing models. We propose a Bayesian statistical model comparison technique which uses k-fold cross-validation across multiple data sets to estimate the likelihood that one model will outperform the other, or that the two will produce practically equivalent results. We use this technique to rank six English part-of-speech taggers across two data sets and three evaluation metrics.