- Number
- 11749265
- Published
- 2023-09-05
- Filed
- 2019-10-04
- Assignee
- DISNEY ENTERPRISES, INC.
- Inventors
- Doggett; Erika Varis et al.
- CPC
- G10L15/22; G10L15/197; G10L15/1822; G10L15/04; G10L15/1815; G10L15/24
- Verdict
- Set aside generic NLU research
- Source
- Google Patents · FreePatentsOnline
Abstract
Various embodiments disclosed herein provide techniques for performing incremental natural language understanding on a natural language understanding (NLU) system. The NLU system acquires a first audio speech segment associated with a user utterance. The NLU system converts the first audio speech segment into a first text segment. The NLU system determines a first intent based on a text string associated with the first text segment, wherein the text string represents a portion of the user utterance. The NLU system generates a first response based on the first intent prior to when the user utterance completes.
Background
BACKGROUND Field of the Various Embodiments (1) The various embodiments relate generally to computer-based language processing technology and, more specifically, to techniques for incremental computer-based natural language understanding. Description of the Related Art (2) Computer-based natural language understanding (NLU) has become a widely popular approach for facilitating human-machine interfaces (HMI) between human users and computer systems. Computer-based NLU is typically included as part of a dialogue system. Such dialogue systems are often deployed in automated chat-based applications, referred to herein as “chatbots,” personal digital assistants, and artificial intelligence (AI) avatars. Typically, a computer-based dialogue system receives audio input from a user via an audio input device, such as a microphone. Notably, the dialogue system waits until an entire spoken utterance from the user has been processed by an automatic speech recognition (ASR) module and received before taking any responsive actions. Such an ASR module can employ various techniques for determining that an entire utterance has been received from a user. For example, the ASR module could wait for a detectable pause in the audio input of a certain duration before concluding that the entirety of a user utterance has been received. The ASR module system then converts the audio input representing the user utterance into text. Then, the dialogue system analyzes the text, and generates an appropriat
Claims
1. A computer-implemented method for performing incremental natural language understanding, the method comprising: acquiring, by a dialogue system that includes at least one processor and a neural network, a first audio speech segment associated with a user utterance; converting the first audio speech segment into a first text segment; executing the neural network to predict a first intent based on a text string associated with the first text segment, wherein the text string represents a portion of the user utterance; determining a confidence score associated with the first intent; modifying at least one marker that is input into the neural network based on at least one of a user associated with the user utterance or a behavior pattern of the user; in response to determining that the confidence score is less than a threshold value, generating, using the neural network with the at least one modified marker, a first response that is unrelated to the first intent prior to when the user utterance completes, wherein the first response indicates that audio associated with the user utterance is being processed and is based on a non-intent specific response library; and in response to determining that the confidence score is greater than the threshold value: predicting an end of the user utterance based on the text string; and generating, using the neural network with the at least one modified marker, a second response based on the end of the user utterance. ||
10. One or more non-transitory computer-readable media storing program instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of: acquiring, by a dialogue system that includes at least one processor and a neural network, a first audio speech segment associated with a user utterance; converting the first audio speech segment into a first text segment; concatenating the first text segment to a text string that represents a portion of the user utterance; executing the neural network to predict a first intent based on the text string; determining a confidence score associated with the first intent; modifying at least one marker that is input into the neural network based on at least one of a user associated with the user utterance or a behavior pattern of the user; in response to determining that the confidence score is less than a threshold value, generating, using the neural network with the at least one modified marker, a first response that is unrelated to the first intent prior to when the user utterance completes, wherein the first response indicates that audio associated with the user utterance is being processed and is based on a non-intent specific response library; and in response to determining that the confidence score is greater than the threshold value: predicting an end of the user utterance based on the text string; and generating, using the neural network with the at least one modified marker, a second response based on the end of the user utterance. ||
18. A system, comprising: one or more memories that include instructions, and one or more processors that are coupled to the one or more memories and, when executing the instructions, perform the steps of: acquiring, by a dialogue system that includes a neural network, an audio speech segment associated with a user utterance; converting the audio speech segment into a text segment; executing the neural network to predict a first intent based on a text string associated with the text segment, wherein the text string represents a portion of the user utterance; determining a confidence score associated with the first intent; modifying at least one marker that is input into the neural network based on at least one of a user associated with the user utterance or a behavior pattern of the user; in response to determining that the confidence score is less than a threshold value, generating, using the neural network with the at least one modified marker, a first response that is unrelated to the first intent prior to when the user utterance completes, wherein the first response indicates that audio associated with the user utterance is being processed and is based on a non-intent specific response library; and in response to determining that the confidence score is greater than the threshold value: predicting an end of the user utterance based on the text string; and generating, using the neural network with the at least one modified marker, a second response based on the end of the user utterance.