The following new features and changes to the service are available.
We released updates to the emotion tone score model. The training dataset was expanded and as a result, it has a much higher precision.
We have released updates to the document emotion tone scores. The new model takes into account emotion profile of sentences, to aggregate the document score.
Emotion Analysis enhancements: Significant enhancements to emotion analysis. There's new training data, a new feature selection process, and augmented lexicons used for words, emojis and slang. This means that there are different but significantly improved emotion scores at the sentence level. There are no input or output changes to the API itself, just more accurate scores.
Targeted Emotion Analysis: The new Targeted Emotion feature allows you to analyze emotions that are associated with specified target phrases in your content.
You can also detect emotions that are associated with entities and keywords with their respective API methods and in a combined call by passing the parameter
Custom models: Now you can tailor AlchemyLanguage to work in domain-specific contexts by using custom models created in Watson Knowledge Studio. These models allow you to customize the output of the Entities API method, and the new Typed Relations API method. Two new custom model management API methods are also now available in AlchemyLanguage.