How to understand the C4.5 algorithm solves the characteristic problem that the ID3 algorithm is biased to the choice value

How to understand the C4.5 algorithm solves the characteristic problem that the ID3 algorithm is biased to the choice value

Consider an extreme case where the value of a property (feature) is so large that each value corresponds to a category with only one. This is based on \[h (D)-H (d| a) \] to know that the value of the next item is 0. The gain in this information will be very large. The C4.5 algorithm adds a penalty item \[h_a (D) =-\sum_{i=1}^n\dfrac{| d_i|} {| d|} \log_2\dfrac{| d_i|} {| d|} \], the greater the value of the penalty if \ (d_i\) is smaller. This solves the problem of the ID3 algorithm.

How to understand the C4.5 algorithm solves the characteristic problem that the ID3 algorithm is biased to the choice value