Table 2. Datasets roles.
| Dataset size (# of rows) | Role | Sentiment analysis methods applied |
|---|---|---|
| 21 K | Majority voting labeling | VADER, GOOG, AMZN, IBM, MSFT |
| 50 K | Majority voting labeling | VADER, GOOG, AMZN, IBM |
| 100 K | Majority voting labeling | VADER, GOOG, AMZN, IBM |
| 250 K | Testing | Majority voting trained algorithms |
Note:
Symbols in the table: VADER, Nltk’s Vader; GOOG, Google Cloud Natural Language’s Analyze Sentiment; AMZN, AWS Amazon Comprehend; IBM, IBM Watson Natural Language Understand, MSFT, Microsoft Azure Text Analytics.