Belief contraction in description logics in a syntax-independent manner is an important and nontrivial problem for ontology management and DL communities. One approach for belief contractions is to take all minimal sets that imply a formula to be contracted and delete one formula from each est, which is also called kernel contraction. In this book I make an investigation of kernel contraction in the description logics EL. I demonstrate that kernel contraction in EL is rational by proving some of its properties. Different ways to generating kernel contractions are presented. I also introduce some novel approaches such as localization and specificity to decide the preference between different certain kernel contractions by exploiting the structure of an EL knowledge base and I give algorithms for computing them.