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