A Thorough Reproducibility Study on Sentiment Classification: Methodology, Experimental Setting, Results

A survey published by Nature in 2016 revealed that more than 70% of researchers failed in their attempt to reproduce another researcher’s experiments, and over 50% failed to reproduce one of their own experiments; a state of affairs that has been termed the ‘reproducibility crisis’ in science. The purpose of this work is to contribute to the field by presenting a reproducibility study of a Natural Language Processing paper about “Language Representation Models for Fine-Grained Sentiment Classification”. A thorough analysis of the methodology, experimental setting, and experimental results are presented, leading to a discussion of the issues and the necessary steps involved in this kind of study.