Classification parameter Sentiment analysis

[DayanaGomez]

Hi!

Sentiment analysis Tool (Text Mining)

The documentation about Max Negative and Min Positive classification parameters explain that can be set any number between -1 and 0 or 0 and 1 respectively. But, when I set one of the parameters in 0, it runs fine one time, after that, if I check the configuration window the parameter that was set to 0 changes to the default value. 

Is any restriction to set and save those parameters with 0?

Thanks!

[PanPP]

Hi @DayanaGomez,

For the sentiment analysis tool, the default value is normally between -1 to 0 and 0 to 1. 

If you are configuring the negative classification, 0 would mean neutral and would default back to a value of -0.1.

This is the default behavior of the tool.

Hope this helps, if it does please like this post and if it helps resolve your problem, mark it as a solution. If you have any other questions, please let us know.

[DayanaGomez]

Thanks @PanPP for your replay.

I understand the idea that 0 means neutral, but I have doubt why it works one time and later it returns to recommended value. If I run the workflow with 0 for max negative, I get different number of records than if I run it with -0.1.

I hope to explain the doubt well.

Thank you!

[martinding]

Hi @DayanaGomez, @PanPP 

I created a dummy dataset and I get what Dayana is saying.

When I set the Max Neg value to 0, it runs fine the first time, but the second time, the tool automatically reverts back to -0.1, therefore giving a different result.

First run

Second run (no change on my part, but tool automatically reverting to -0.1)

I think what PanPP is saying is that 0 is always neutral from the tool's perspective (by design), so it always reverts to something negative (in this case -0.1) if the tool detects that it's been set to neutral.

I guess if you want to force the classification to be negative when sentiment is neutral, you could use a formula tool to label based on the raw sentiment scores, rather than outputting the label using the Sentiment tool.

[DayanaGomez]

Thanks for your contributions.

@martinding, this example represents the situation well. Now I understand, as PanPP mentioned, this is the default behavior of the tool, so if the parameter is set to 0, the tool defaults back to a non-neutral value.

[PanPP]

So the reason why this is the default behavior is because these are the standard thresholds recommended for the Vader algorithm (which is used under the hood in this case).

Hope this helps.