There is a need for live meeting support and remote assitants which should also be able to retrieve historical data e.g. in case a person joins a meeting late. The hub is a Java-based system for live distribution of data between recognition systems and browsers. On one side, there are recognition systems (ASR, gesture recognition) that detect patterns (e.g. events, words) and produce annotations. On the other side there are browsers waiting for multimodal data. Traditional database architecture have a main disadvantage when dealing with real time data: it is difficult for the client browsers to know when the database has been updated without e.g. polling it. Using this proposed scheme, recognizers produce data which is sent to a main process server: the Hub. Then other processes (consumer clients, browsers) happen to register interest with the hub.