search

openai / generating-reviews-discovering-sentiment

Code for "Learning to Generate Reviews and Discovering Sentiment"
https://arxiv.org/abs/1704.01444
MIT License
1.51k stars 379 forks source link

Is neutral prediction possible? #34

Closed itisjoshi closed 7 years ago

itisjoshi commented 7 years ago

We are trying to build sentiment analysis for email conversations. The problem we are facing is most of the emails were neutral conversations.

This models predicts neutral sentiment as positive or negative.

My questions is there any specific range for negative, positive and neutral sentiments?

Is it possible to predict sentiments as neutral?

ahirner commented 7 years ago

The typical ranges for sentiments inferred or generated are discussed in #21.

Newmu commented 7 years ago

One of the tasks the paper benchmarks on is the very negative, negative, neutral, positive, very positive version of the Stanford Sentiment Treebank and it performs close to state of the art even though it includes neutral suggesting there's no fundamental issue with neutral.

The more likely problem is out-of-domain issues. The model is trained on product reviews which can have pretty different language from an email. There's no inherent "range" but you could probably fit simple thresholds for your task given only a few labeled negative, neutral, and positive examples.