The voice user interface is the primary interface to the voice portal for the end users. As mentioned previously, flexibility is build into the voice portal through adaptation on the user interface to local languages, speaking patterns, syntax, and cultural norms. However, for implementation efficiency, the underlying applications—such as voice-activated dialing—should be independent of language and culture. Therefore, a layer known as the context application programming interface (CPI) exists between the VUI and the applications that translates the utterances received by the VUI into a representation that can be used by any application, regardless of language. Once the subscriber has been identified, the VUI begins the conversation by playing a prompt and listening for input from the caller. Speech input is directed to the speech-recognition engine, which analyzes the speech based on the grammar specified for the caller, while DTMF (dual-tone multifrequency, or keypad) tones are analyzed by the telephony hardware. This input is returned to the VUI, which passes the input tokens based on the utterances and the DTMF tones to the CPI. The CPI determines the appropriate application or context to handle the input, and then issues a series of actions to gather the information requested by the user. This information is then passed back to the VUI for presentation. This presentation could make use of prerecorded prompts, audio files, real-time streaming of audio from external sources (such as the Internet), or generated speech from a TTS engine. The platform then sends the data over the telephone interface.