A typical Spring MVC application will have two contexts: the root context and the servlet context.
The root context is loaded by the ContextLoaderListener. The ContextLoaderListener loads the ApplicationContext and sets it as an attribute in the ServletContext. This makes it available to other Servlet, Filter, and XxxListener instances.
The servlet context is another ApplicationContext loaded by the DispatcherServlet when its init() method is called. Since the init() method of a Servlet is called after the contextInitialized() method of a ServletContextListener is called, the DispatcherServlet has access to the root context set as an attribute in the ContextLoaderListener and can therefore use the beans declared there.
[...] creates the beans defined there, overriding the definitions of any
beans defined with the same name in the global scope.
A bean definition is, for example,
<bean id="someBean" class="com.company.SomeBean">
<property name="someProp" value="some value" />
</bean>
Spring creates a BeanDefinition object for this and all other <bean> declarations (or @Bean annotation methods). When the ApplicationContext is refreshed, Spring generates actual instances from these bean definitions.
What the above quote from the documentation is saying is that if you have a <bean> declaration with the same name or id in both the root context and the servlet context, the servlet context one will overwrite the root context one.