Sentiment analysis examines text using natural language analysis techniques
and attempts to determine the overall sentiment of the block of text.
TripleBlind supports several of the most popular systems for performing
this type of analysis.

Sentiment can be expressed as a number between -1.0 and 1.0, where a negative
value indicates the text has a highly negative tone, 1.0 indicates a very
positive tone, and 0.0 is totally neutral.  For example, the text "I hated my
meal" might return a sentiment of -0.85, "Service was ok" might be 0.25, and no
text or "We ate at 7:00" might be 0.0.

==== BERT Analysis ====

The BERT_Fine_Tune demonstration uses Google's Bidirectional Encoder
Representations from Transformers (BERT) library to train on top of their
default English language model with custom text.  In this case a movie review
database is used, training on textual reviews and their associated movie
rating.

The custom model created by this training is next used to perform inference
on text blocks to predict a review rating from the text alone.

NOTE: The BERT library is computationally intensive, so training on an Access
Point with GPUs is recommended.
