A Comparison of Tools for Sentiment Polarity over Code Review Comments #2
Description
The dataset to be used is the gold standard code review dataset published at http://amiangshu.com/papers/senticr-ase.pdf This comprises 1600 code review comments manually labeled.
We'll be limited to the sentiment polarity labels used in that dataset: negative, neutral, positive.
Based on the following benchmark study (https://arxiv.org/pdf/1803.06525.pdf), SentiCR performed better for code review comments compared to SentiStrength, SentiStrengh-SE, and Senti4SD.
Here, we're going to compare SentiCR with IBM Tone Analyzer, EMTk, and possibly others if time permits (Google, Microsoft, AWS).