Abstract: Computer-implemented techniques for answer time prediction in a contact center service of a provider network. While a delayed processing timing has not been met, a set of contact queuing context-actual answer time data for a set of contact inquiries serviced is received as a first set of contact queuing context-actual answer time data. When the delayed processing timing has been met, a new queuing model is learned based on the first set of contact queuing context-actual answer time data and a previous set of contact queuing context-actual answer time data for a previous set of contact inquiries serviced or a previous version of the queuing model. A request to predict an answer time for a target contact queuing context is received and an answer time for the target contact queuing context is predicted based on the new queuing model. The predicted answer time is provided to a contact via a contact communications channel.

Abstract: A system (140) aids in the tuning of one or more speech applications. The system (140) may receive event data associated with multiple user interactions with one or more speech applications and store the event data in a database (310). In a real time mode, the system (140) may receive a request for information, retrieve event data from the database (310) based on the request, formulate a response to the request using the retrieved event data, and present the response to aid in the tuning of the one or more speech applications. In a periodic mode, the system (140) may periodically analyze the event data to identify potential problem areas, generate results of the periodic analyzing, and present the results to aid in the tuning of the one or more speech applications.