Why and how does the log transform reduce skewness in a dataset to make it resemble normal distribution?

I am studying a statistics book. At some point when working with a dataset, the book applies log-transform to one of the columns because the data is skewed and we want it to be as close to a bell curve (normal distribution) as possible. I understand the reason for using some transform to do this, this is basically scaling. I just don't really understand why the log-transform does this. How exactly does it do this, from a mathematical stand point?

Is this because the normal distribution's density function is in the format of e to the power of something, so it's like a reverse of log natural base? So if we want to get to the normal distribution we use log? Or am I totally off?

This isn't part of homework or anything. The book is mostly coding, so it offers nothing more about the theoretical stuff. This is just me being curious.

New contributor
samir is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
• Skewed data can sometimes be approximated using a log-normal distribution, which has the property that if $X$ is log-normally distributed, then $Y=\ln(X)$ is normally distributed. Jan 15 at 0:13