I don't even know what my problem is called, so I can't research literature on it. I'm hoping someone can point me in the right direction. I have a little experience with sentiment analysis, but what I want to do is more specific. Rather than evaluating the general sentiment of a body of text, I want to evaluate the sentiment about a particular subject in a body of text. For example, let's say I have an article about President Obama. One sentence says

President Obama arrived in Florida and saw the horrific destruction left behind by the hurricane.

While the general / overall sentiment of this sentence would like be negative, Obama is spoken about fairly neutrally. This is what I mean by about-sentiment. I want to evaluate (with a computer) how positively or negatively the text speaks about a specific subject.

What is this called? Is there literature on it? How can I proceed?

1 Answer 1


Google NLP and others call this 'Entity Sentiment Analysis' (versus document sentiment analysis). Also see "aspect based sentiment', e.g., https://arxiv.org/pdf/1805.07043.pdf Finally, look at OpenAI's approach. It isn't entity/aspect oriented but the sentiment shifts as it reads: https://blog.openai.com/unsupervised-sentiment-neuron/


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.