TAPAGE (edition de Tableaux par la Parole et la Geste) is a synergistic pen/voice system for designing and correcting tables.
To capture signals emitted during a user's interaction, TAPAGE integrates a set of modality agents, each responsible for a very specialized kind of signal [[9]]. The modality agents are connected to an ``interpret agent'' which is responsible for combining the inputs across all modalities to form a valid command for the application. The interpret agent receives filtered results from the modality agents, sorts the information into the correct fields, performs type-checking on the arguments, and prompts the user for any missing information, according to the model of the interaction. The interpret agent is also responsible for merging the data streams sent by the modality agents, and for resolving ambiguities among them, based on its knowledge of the application's internal state. Another function of the interpret agent is to produce reflexes: reflexes are actions output at the interface level without involving the functional core of the application.
The TAPAGE system can accept multimodal input, but it is not a distributed system; its functional core is fixed. In TAPAGE, the set of linguistic input is limited to a verb object argument format.