A speech-enabled internet based computing system includes a configurable speech recognition engine used for interacting with content on a web accessible page. The speech recognition engine is distributed across a client and server architecture, and is adaptive so that speech processing operations can be allocated as needed between the two. This allows for support for client devices having differing computing capabilities. Natural language operations can also be supported as desired. A user can thus interact with a web page and select items of interest using speech as a mode of input. Dynamic grammars can assist in the recognition operations to improve speed and comprehension.