Timeline for What do we mean as Positive or Negative in Sentiment Analysis?
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 7, 2021 at 20:58 | comment | added | n1k31t4 |
A model like BERT trained on real world data will map and interpolate the sentiment from the labels with which it was trained. That means it will likely know that terrorist has a negative connotation and that good adds a positive sentiment to the noun it modifies. As good terrorist is a rare combination (on its own, without more context), it is hard to know how a learned model like BERT would rate its sentiment, given the juxtaposition.
|
|
| Sep 7, 2021 at 19:55 | comment | added | Mahdi Amrollahi | So, you mean that it doesn’t depend on the subject? For example if I say “ terrorists are good” and “ students are good”, they evaluated the same? I am actually taking about nn based like BERT not something like VADER. | |
| Sep 7, 2021 at 18:33 | history | answered | Brian Spiering | CC BY-SA 4.0 |