Spoken Language of the Deployment Region
Because speech recognizers are typically fine-tuned to a particular language, regional differences in spoken languages affect the recognition rate of those recognizers. Conversely, these same users may have difficulty understanding the text-to-speech synthesis rendered by similarly designed TTS engines.
Using the spoken language as one of the marketing criteria for the proposed application, the designer should consider the majority language. In any case, the designer should be prepared to do preliminary testing to determine the speech recognition rates and the speech synthesis intelligibility before too much progress is made on further application design. The application designer should determine the language spoken by a majority of the deployment region/country. The designer should then look for voice portal providers or hosts that support the desired language on both the automatic speech recognition (ASR) and text-to-speech (TTS) engines. Where a multiple language interface is required, not only should the chosen languages still reflect the deployment region, but the interface design will need to consider usage by multilingual speakers.