What is the sentiment accuracy of your system “out-of-the-box”?
The sentiment accuracy “out-of-the-box” depends on the nature of the processed content. Shorter content, such as tweets or SMS messages, do not carry enough grammatical and syntactical information for the engine to invoke higher-level analysis. Thus, we typically see 60-65% accuracy on shorter content out-of-the-box, meaning according to the given category samples Semantria’s predictions are 60-65% accurate. Longer form content usually gives us 70-75% accuracy out of the box because there is more content and context for Semantria’s engine to work from.
To improve sentiment accuracy see the Sentiment Adjustment article or to understand Semantria's Sentiment Methodology, see the Sentiment Analysis article.