Abstract: A speech interface device is configured to perform natural language understanding (NLU) processing in a manner that optimizes the use of resources on the speech interface device. In an example process, a domain classifier(s) is used to generate domain classifier scores associated with multiple candidate domains, and the candidate domains can then be evaluated, one candidate domain at a time, in accordance with the domain classifier scores (e.g., starting with a highest scoring candidate domain). For each candidate domain undergoing the evaluation, input data is by that domain's NLU model(s), and, as soon as a domain-specific NLU model(s) produces a NLU result with a confidence score that satisfies a threshold confidence score, the evaluation can be stopped for any remaining candidate domains.

Abstract: A speech interface device is configured to process user speech by storing, in volatile memory of the speech interface device, audio data that represents user speech, and inputting first audio data, of the stored audio data, to an automatic speech recognition (ASR) component of the speech interface device, determining that a criterion is satisfied, and, based on the criterion being satisfied, maintaining second audio data in the volatile memory. The ASR component may generate text data based on the first audio data, a natural language understanding (NLU) component of the speech interface device may generate NLU data based on the text data, and, if the NLU data corresponds to a recognized intent, the second audio data may be deleted. Otherwise, speech processing can be resumed by inputting the second audio data to the ASR component.