Although an MMI description language XISL1.1 (eXtensible Interaction Scenario Language) developed for web-based MMI applications has some desirable features such as high extensibility and controllability of modalities, it has some issues concerning capabilities for describing slot-filling style dialogs, a large amount of description for modality combination of inputs, and complicated authoring when handling XML contents. In this paper we provide a new version of XISL (XISL2.0) that resolves these issues by using VoiceXML-based syntax, modality-independent input description, and XForms-like data model. The results of comparison between XISL1.1 and XISL2.0 showed that XISL2.0 reduced the amount of description by 40% as compared with XISL1.1, and XISL2.0 succeeded to mitigate the difficulties in the authoring work of developing MMI applications.