CoreNLP's sentiment analysis is too error-prone, it thinks most neutral sentences are negative.
Perhaps I can implement a separate annotator that mimics the tree output or perhaps make a container for both the tree and other types of sentiment analysis outputs. For example, SentimentContext could be a container class for sentiment content with generics, where T would be the Tree class for CoreNLP sentiment analysis. The container could contain both local and sentence sentiment, simplying matters a bit.
I want to implement the first (and perhaps only) alternative sentiment analysis system using SentiStrength.
CoreNLP's sentiment analysis is too error-prone, it thinks most neutral sentences are negative.
Perhaps I can implement a separate annotator that mimics the tree output or perhaps make a container for both the tree and other types of sentiment analysis outputs. For example, SentimentContext could be a container class for sentiment content with generics, where T would be the Tree class for CoreNLP sentiment analysis. The container could contain both local and sentence sentiment, simplying matters a bit.
I want to implement the first (and perhaps only) alternative sentiment analysis system using SentiStrength.