Abstract: Methods and apparatus for performing speech recognition using a garbage model. The method comprises receiving audio comprising speech and processing at least some of the speech using a garbage model to produce a garbage speech recognition result. The garbage model includes a plurality of sub-words, each of which corresponds to a possible combination of phonemes in a particular language.

Abstract: Methods and apparatus for performing speech recognition using a garbage model. The method comprises receiving audio comprising speech and processing at least some of the speech using a garbage model to produce a garbage speech recognition result. The garbage model includes a plurality of sub-words, each of which corresponds to a possible combination of phonemes in a particular language.

Abstract: Aspects of the disclosure are directed to natural language processing. An input interface of a computing device receives input (e.g., speech input) and generates a digital signal corresponding to that input. Text corresponding to the digital signal is obtained, and the text is processed using each of a context-free and a context-specific linguistic model to generate linguistic processing results for that text. The text and linguistic processing results may be processed using a NLU model to generate an NLU recognition result corresponding to the input received at the input interface. The text and the linguistic processing results may also be annotated and used to train a NLU model. The linguistic processing results may relate to, e.g., the tokenization of portions of the text, the normalization of portions of the text, sequences of normalizations for portions of the text, and rankings and prioritization of the linguistic processing results.

Abstract: Embodiments disclosed herein enable scaling up and making advanced natural language (NLU) applications more robust. According to one embodiment, state(s) associated with a dialog session may be recorded to a non-transitory medium. The dialog session may be suspended after a given period of inactivity and later automatically awakened based on unique client, session, or device identifier, or any combination thereof. Memory and resources associated with the suspended session may be reclaimed, the memory and resources being otherwise held by the session during the period of inactivity, enabling higher density (e.g., a larger number of sessions supported). Embodiments disclosed herein obviate a need for sticky dialog sessions, enabling higher density, and may further failover protection and fault tolerance for the dialog sessions.

Abstract: A method for conducting a call between a caller and an interactive voice response (IVR) system, the caller using a device to conduct the call, the device configured to execute a virtual assistant, the method comprising using the virtual assistant to conduct the call at least in part by influencing the style of information provided to the caller during the call and/or the content of information passed between the device and the IVR system during the call.

Abstract: Embodiments disclosed herein enable scaling up and making advanced natural language (NLU) applications more robust. According to one embodiment, state(s) associated with a dialog session may be recorded to a fixed medium. The dialog session may be suspended after a given period of inactivity and later automatically awakened based on unique client, session, or device identifier, or any combination thereof. Memory and resources associated with the suspended session may be reclaimed, the memory and resources being otherwise held by the session during the period of inactivity, enabling higher density (e.g., a larger number of sessions supported). Embodiments disclosed herein obviate a need for sticky dialog sessions, enabling higher density, and may further failover protection and fault tolerance for the dialog sessions.

Abstract: A speech server and methods provide audio stream analysis for tone detection in addition to speech recognition to implement an accurate and efficient answering machine detection strategy. By performing both tone detection and speech recognition in a single component, such as the speech server, the number of components for digital signal processing may be reduced. The speech server communicates tone events detected at the telephony level and enables voice applications to detect tone events consistently and provide consistent support and accuracy of both inbound and outbound voice applications independent of the hardware or geographical location of the telephony network. In addition, an improved opportunity for signaling of an appropriate moment for an application to leave a message is provided, thereby supporting automation.