Text Analytics, as part of Azure Cognitive Services, has released the latest version of its sentiment analysis capability. It supports English and is available in three regions Central Canada, East Asia, and Central US. Additional languages and regions will be added in subsequent releases.
The new model has the following improvements:
- Improved accuracy: Significant improvement in detecting positive, neutral, negative, and mixed sentiment in text documents over previous versions.
- Document and Sentence-level Sentiment Score: Detect the sentiment of both a document and its individual sentences. If the document includes multiple sentences, each sentence is also assigned a sentiment score.
- Sentiment category and score: The API now returns sentiment categories (positive, negative, neutral and mixed) for text, in addition to a sentiment score.
- Improved output: Sentiment analysis now returns information for both an entire text document, and its individual sentences.
The API's request interface remains the same. The input is identical to previous version, but the output provides additional information listed above. For more information, see the documentation.