Granted Patent B2

US 12,494,205 · App. 18/094,284 · Granted Dec 9, 2025

Inventors: Taylor G. Smith (Allen, TX); Tori Salido (Austin, TX); Sebastin Ajay John Jeyaseelan (Thoraipakkam, IN); Kordel K. France (Allen, TX)

Assignee: TOYOTA CONNECTED NORTH AMERICA, INC.

G10L15/26G06F40/166G10L25/78

Abstract

Claims (40)

1 . An apparatus comprising:

a receiver configured to receive utterances from a user via an input device; and

a processor configured to

identify a plurality of sentences spoken by the user from the utterances;

convert the utterances into a sequence of tokens and storing the sequence of tokens within a data structure, wherein the storing comprises adding padding tokens to the data structure in between the sequence of tokens to normalize a structure among the plurality of sentences within the data structure;

execute a machine learning model on the data structure with the added padding tokens to determine to make a prediction;

remove the padding tokens from the data structure; and

execute a natural language processing (NLP) model on the sequence of tokens within the data structure with the padding tokens removed to determine a response to the user,

wherein a first word in an input utterance and a last word in the input utterance are identified, wherein a token that corresponds to the first word and a token that corresponds to the last word are stored in different slots of the data structure.

2 . The apparatus of claim 1 , wherein the processor is configured to store one or more padding tokens between two sub-sequences of tokens that correspond to two sentences uttered by the user during a conversation to normalize a size of the two sub-sequences of tokens within the data structure.

3 . The apparatus of claim 1 , wherein the data structure comprises a tensor with slots that hold data and the processor is configured to store each sentence as a sequence of tokens within a sequence of slots of the tensor with one or more padding tokens inserted in slots between each sequence of tokens within the tensor.

4 . The apparatus of claim 1 , wherein the processor is configured to move the sequence of tokens within the data structure to collapse the sequence of tokens after the padding tokens have been removed and execute the NLP model on the collapsed sequence of tokens in the data structure.

5 . The apparatus of claim 1 , wherein the machine learning model comprises a sequence model that receives the data structure as an input and determines whether to control a system to take an action or to request more utterances from the user.

6 . The apparatus of claim 1 , wherein the processor is configured to insert a padding token into a slot of the data structure immediately before the token that corresponds to the first word and after a token that corresponds to a last word in a previous input utterance.

7 . The apparatus of claim 1 , wherein the processor is configured to insert a padding token into a slot of the data structure immediately after the token that corresponds to the last word in the input utterance and before a token that corresponds to a first word in a next input utterance.

8 . A method comprising:

receiving utterances from a user via an input device;

identifying a plurality of sentences spoken by the user from the utterances;

converting the utterances into a sequence of tokens and storing the sequence of tokens within a data structure, wherein the storing comprises adding padding tokens to the data structure in between the sequence of tokens to normalize a structure among the plurality of sentences within the data structure;

executing a machine learning model on the data structure with the added padding tokens to determine to make a prediction;

removing the padding tokens from the data structure; and

executing a natural language processing (NLP) model on the sequence of tokens within the data structure with the padding tokens removed to determine a response to the user,

wherein a first word in an input utterance and a last word in the input utterance are identified, wherein a token that corresponds to the first word and a token that corresponds to the last word are stored in different slots of the data structure.

9 . The method of claim 8 , wherein the storing comprises inserting one or more padding tokens between two sub-sequences of tokens corresponding to two sentences uttered by the user to normalize a size of the two sub-sequences of tokens within the data structure.

10 . The method of claim 8 , wherein the data structure comprises a tensor with slots that hold data and the storing comprises storing each sentence as a sequence of tokens within a sequence of slots of the tensor with one or more padding tokens inserted in slots between each sequence of tokens within the tensor.

11 . The method of claim 8 , wherein the removing further comprises collapsing the sequence of tokens within the data structure to concatenate the sequence of tokens after the padding tokens have been removed and executing the NLP model on the collapsed sequence of tokens in the data structure.

12 . The method of claim 8 , wherein the machine learning model comprises a sequence model that receives the data structure as an input and determines whether to control a system to take an action or to request more utterances from the user.

13 . The method of claim 8 , wherein the converting comprises inserting a padding token into a slot of the data structure immediately before the token corresponding to the first word and after a token corresponding to a last word in a previous input utterance.

14 . The method of claim 8 , wherein the converting comprises inserting a padding token into a slot of the data structure immediately after the token corresponding to the last word in the input utterance and before a token corresponding to a first word in a next input utterance.

15 . A non-transitory computer-readable storage medium comprising instructions, that when read by a processor, cause the processor to perform a method comprising:

receiving utterances from a user via an input device;

identifying a plurality of sentences spoken by the user from the utterances;

converting the utterances into a sequence of tokens and storing the sequence of tokens within a data structure, wherein the storing comprises adding padding tokens to the data structure in between the sequence of tokens to normalize a structure among the plurality of sentences within the data structure;

executing a machine learning model on the data structure with the added padding tokens to determine to make a prediction;

removing the padding tokens from the data structure; and

executing a natural language processing (NLP) model on the sequence of tokens within the data structure with the padding tokens removed to determine a response to the user,

wherein a first word in an input utterance and a last word in the input utterance are identified, wherein a token that corresponds to the first word and a token that corresponds to the last word are stored in different slots of the data structure.

16 . The non-transitory computer-readable storage medium of claim 15 , wherein the storing comprises inserting one or more padding tokens between two sub-sequences of tokens corresponding to two sentences uttered by the user to normalize a size of the two sub-sequences of tokens within the data structure.

17 . The non-transitory computer-readable storage medium of claim 15 , wherein the data structure comprises a tensor with slots that hold data and the storing comprises storing each sentence as a sequence of tokens within a sequence of slots of the tensor with one or more padding tokens inserted in slots between each sequence of tokens within the tensor.

18 . The non-transitory computer-readable storage medium of claim 15 , wherein the removing further comprises moving the sequence of tokens within the data structure to collapse the sequence of tokens after the padding tokens have been removed and executing the NLP model on the collapsed sequence of tokens in the data structure.

Assignments (1)

Continuity (1)

References Cited (31)