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).
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).