Can't get sample to report non-toxic sentiment

[dominiqueplante]

Problem encountered on https://dotnet.microsoft.com/learn/ml-dotnet/get-started-tutorial/consume Operating System: macos

I changed to different values including

input.SentimentText = "That is good.";

and

input.SentimentText = "Get yourself some help.";

(this is a line that was non-toxic in the CSV)

I was trying to find a value that gives mem non-toxic sentiment

[CESARDELATORRE]

@briacht - Can you check this issue?

[ronnyek]

Sentiment analysis models definitely seem to be... not well trained, or not very effective.

I am consuming https://github.com/dotnet/samples/raw/main/machine-learning/models/sentimentanalysis/sentiment_model.zip in an api example... and plugging in values I almost always get the wrong prediction values out. I have seen a prediction false, so I know prediction engine is doing SOMETHING

"I hate this place", "this place sucks", "my least favorite place", "I dislike this place" all return with something like

{
	"prediction": true,
	"probability": 0.8202787,
	"score": 1.5182369
}