The semantic gap characterizes the difference between two descriptions of an object by different linguistic representations, for instance languages or symbols. The semantic gap can be defined as "the difference in meaning between constructs formed within different representation systems".
I need to find a clear answers to the second part
Any ML model assumes a particular representation of the data, and connects bits of information only within this representation.
For example a linear regression model assumes a linear relation between the features and the target variable. A Naive Bayes model assumes that the features are independent of each other. And these are only the most obvious kind of simplifications made by ML models.
Naturally this results in different models representing data differently. Any semantic interpretation based on some model outcome is potentially biased by the assumptions made by the model.
