18526991. UTTERANCE CLASSIFIER simplified abstract (Google LLC)

UTTERANCE CLASSIFIER

Organization Name

Google LLC

Inventor(s)

Nathan David Howard of Mountain View CA (US)

Gabor Simko of Santa Clara CA (US)

Maria Carolina Parada San Martin of Boulder CO (US)

Ramkarthik Kalyanasundaram of Cupertino CA (US)

Guru Prakash Arumugam of Sunnyvale CA (US)

Srinivas Vasudevan of Mountain View CA (US)

UTTERANCE CLASSIFIER - A simplified explanation of the abstract

This abstract first appeared for US patent application 18526991 titled 'UTTERANCE CLASSIFIER

Simplified Explanation

The method described in the abstract involves using a neural network-based utterance classifier to determine if a spoken utterance is directed towards an automated assistant server, and generating a response if it is.

• Neural network-based utterance classifier with LSTM layers
• Textual representation generated for each word of spoken utterance
• Classifier trained on negative training examples
• Determines if spoken utterance is directed towards automated assistant server
• Generates response if utterance is directed towards server

Potential Applications

This technology could be applied in various industries such as customer service, virtual assistants, and voice-controlled devices.

Problems Solved

This technology helps in accurately identifying spoken utterances directed towards automated assistant servers, enabling prompt and relevant responses to user queries.

Benefits

- Improved user experience with automated assistant servers - Efficient handling of spoken queries - Enhanced accuracy in understanding user intent

Potential Commercial Applications

Optimizing customer service interactions, enhancing virtual assistant capabilities, and improving voice-controlled devices are potential commercial applications of this technology.

Possible Prior Art

Prior art in this field may include similar neural network-based classifiers for speech recognition and natural language processing tasks.

How does this technology handle different languages or accents in spoken utterances?

The abstract does not provide information on how the neural network-based utterance classifier deals with variations in language or accents that may affect the accuracy of determining if an utterance is directed towards the automated assistant server.

What is the computational efficiency of the method described in the abstract?

The abstract does not mention anything about the computational efficiency of the method, such as the processing time required to generate responses to spoken utterances.

Original Abstract Submitted

A method includes receiving a spoken utterance that includes a plurality of words, and generating, using a neural network-based utterance classifier comprising a stack of multiple Long-Short Term Memory (LSTM) layers, a respective textual representation for each word of the of the plurality of words of the spoken utterance. The neural network-based utterance classifier trained on negative training examples of spoken utterances not directed toward an automated assistant server. The method further including determining, using the respective textual representation generated for each word of the plurality of words of the spoken utterance, that the spoken utterance is one of directed toward the automated assistant server or not directed toward the automated assistant server, and when the spoken utterance is directed toward the automated assistant server, generating instructions that cause the automated assistant server to generate a response to the spoken utterance.