In this paper we describe a solution to make Web pages more suitable for vocal browsing by analyzing and modifying their logical structure. The solution exploits intermediate logical descriptions that are automatically created by reverse engineering techniques. The adaptation engine aims to identify the main logical structure of the Web page components and remove the aspects specific to the graphical modality. Then, a vocal implementation is generated to support browsing, which begins by the user's selecting from the main components. It is possible to customize some parameters of the adaptation transformation in order to better control its results.