SentimentSnatcher

SentimentSnatcher is a mechanism that combines the abilities of a sentiment classification and text summarization model to perform sentiment analysis in the form AI-generated text. I fine-tuned a Bert model to perform sentiment classification on natural language. To perform text summarization, I fine-tuned a Bart Seq2Seq model on a dataset of dialogue transcriptions and their relevant summaries.

The end result was a container "model" (i.e. SentimentSnatcher) comprised of both the sentiment classification and generative summarization language model. SentimentSnatcher is able to take a group of up to 100 reviews and generate a single summary that captures the overall sentiment of the reviews fed into it.