Is there an objective benchmark to measure the performance of voice assistants? Usually, NLP tasks have ways to measure their performance.