The Language service has improved the precision of pretrained models and released the ability to do sentence and
document level sentiment.  API improvements allow for batch-scoring, allowing you to score up to 100 documents
in a single request.  

The Language service documentation shows you how to use this new service.