Easily find the sentiment of a sentence targeted for social media and support messages.
- Support Routing: route inbound messages to the proper support team.
- Have a dedicated crisis response team for negative messages
- Use chat bots or automated replies for neutral messages
- Discover extremely positive messages as showcases of happy customers
- Filter extremely negative messages in chat rooms
The input should contain an array of texts.
{
"texts": [
"The book was good.",
"At least it isn't a horrible book.",
"The book was only kind of good.",
"Today SUX!",
"John is not smart, handsome, nor funny.",
"The plot was good, but the characters are uncompelling and the dialog is not great."
]
}
The output will map each input message to a score. Positive texts receive positive scores while negative texts receive negative scores.
{
"texts": {
"The book was good.": 0.4404,
"At least it isn't a horrible book.": 0.431,
"The book was only kind of good.": 0.3832,
"Today SUX!": -0.5461,
"John is not smart, handsome, nor funny.": -0.7424,
"The plot was good, but the characters are uncompelling and the dialog is not great.": -0.7042
}
}
The underlying model was designed by a group at Georgia Institute of Technology and is "specifically attuned to sentiments expressed in social media".
See the MAINTAINERS.md to learn more about the underlying model as well as how to contribute to the app.