Sentiment extraction is the process of inferring user sentiment (positive, negative, or neutral) from text (typically call center logs, forums, and social media).
The sentiment extraction functions are:
- TrainSentimentExtractor, which trains a model—takes training documents and outputs a maximum entropy classification model
- ExtractSentiment, which uses either the classification model or a dictionary model to extract the sentiment of each input document or sentence; that is, to output predictions
- EvaluateSentimentExtractor, which uses test data to evaluate the precision and recall of the predictions
The sentiment extraction functions support English, Simplified Chinese, and Traditional Chinese text.