Voice and textual interface for closed-domain environment
AssigneeEndgame, Inc.
InventorsRobert Filar, Richard Seymour, Alexander Kahan
Life of this patent
Estimating the term…
Abstract
An improved system and method is disclosed for receiving a spoken or written utterance, identifying and replacing certain words within the utterance with labels to generate a simplified text string representing the utterance, performing intent classification based on the simplified text string, and performing an action based on the intent classification and the original words that were replaced.
Claims24
A method of providing a conversational interface for a computing device, the method comprising:
- receiving, by the computing device, an utterance from a user;
- tokenizing, by the computing device, the utterance into a tokenized utterance;
- generating, by the computing device, feature vectors for the tokenized utterance, the feature vectors comprising a plurality of terms and a value for each of the plurality of terms indicating the presence of each of the plurality of terms in the tokenized utterance;
- identifying, by the computing device using the tokenized utterance and the feature vectors, select terms in the tokenized utterance to replace with labels;
- replacing, by the computing device, the select terms with labels to generate redacted text;
- tokenizing, by the computing device, the redacted text into tokenized redacted text;
- determining, by the computing device, an intent of the utterance using the tokenized redacted text;
- performing, by the computing device, an action based on the intent.
A computing device comprising a processor, memory, and non-volatile storage, the non-volatile storage containing a computer program comprising instructions for performing the following steps when executed by the processor:
- receiving an utterance from a user;
- tokenizing the utterance into a tokenized utterance;
- generating feature vectors for the tokenized utterance, the feature vectors comprising a plurality of terms and a value for each of the plurality of terms indicating the presence of each of the plurality of terms in the tokenized utterance;
- identifying, using the tokenized utterance and the feature vectors, select terms in the tokenized utterance to replace with labels;
- replacing the select terms with labels to generate redacted text;
- tokenizing the redacted text into tokenized redacted text;
- determining an intent of the utterance using the tokenized redacted text;
- performing, by the computing device, an action based on the intent.
Dependencies are inferred from the claim text.