Granted Patent B2

US 12,499,890 · App. 18/319,729 · Granted Dec 16, 2025

Inventors: Durgesh Kumar (Bangalore, IN); Saurabh Tahiliani (Noida, IN)

Assignee: Verizon Patent and Licensing Inc.

G10L15/22G10L13/02G10L15/02G10L15/063G10L15/1815G10L15/183G10L25/63

Abstract

Claims (98)

1 . A method, comprising:

receiving a transcript of a current communication with a customer at a geo-locale;

retrieving a first set of transcripts of historic communications associated with the customer and a second set of transcripts of historic communications that occurred at the geo-locale;

computing a prompt input vector based on

a personalized contextual vector obtained based on the transcript and the first set of transcripts, wherein the personalized contextual vector represents personal characteristics of the customer exhibited in the current and the historic communications,

a geo-localized contextual vector obtained based on the second set of transcripts, wherein the geo-localized contextual vector represents characteristics of the geo-locale;

computing a current text vector based on the transcript to characterize the current communication;

generating, by a language model, a prompt based on the prompt input vector and the current text vector, wherein the language model is pretrained via deep learning to predict a next sentence in the current communication based on contexts of the current communication and historic communications captured by the current text vector and the prompt input vector, respectively, and wherein the language model learns, via training, to predict in different contexts;

creating an interactive voice response (IVR) communication based on the prompt for responding to the customer; and

sending the IVR communication to the customer.

2 . The method of claim 1 , wherein the computing the prompt input vector comprises:

processing the transcript and the first set of transcripts to generate a first processed result;

estimating a first plurality of features characterizing the customer based on the first processed result; and

computing the personalized contextual vector based on the first plurality of features and personalized embeddings previously trained via machine learning.

3 . The method of claim 2 , wherein the first plurality of features include at least one of:

a first level intent of the customer estimated based on each phrase identified in the first processed result;

a second level intent of the customer estimated based on each session of communication transcript identified in the first processed result;

an emotional state of the customer estimated based on at least one of the first level intent, the second level intent, and the first processed result; and

at least one unanswered question detected from the first processed result.

4 . The method of claim 2 , further comprising:

processing the second set of transcripts to generate a second processed result;

estimating a second plurality of features characterizing the geo-locale based on the second processed result;

computing the geo-localized contextual vector based on the second plurality of features and geo-localized embeddings previously trained via machine learning;

generating the prompt input vector by combining the personalized contextual vector and the geo-localized contextual vector.

5 . The method of claim 4 , wherein the second plurality of features include at least one of:

at least one geo-localized intent estimated with respect to the geo-locale based on the second processed result;

at least one intent-based event level detected based on the second processed result, wherein each of the at least one intent-based event level characterizes a corresponding one of the at least one geo-localized intent; and

an emotional state associated with the geo-locale estimated based on the second processed result.

6 . The method of claim 1 , wherein the creating an interactive voice response (IVR) communication based on the prompt comprises:

generating an IVR communication in a textual form based on the prompt;

converting the IVR communication in the textual form into a voice signal based on text-to-speech transformation; and

outputting the voice signal as the IVR communication.

7 . A machine readable and non-transitory medium having information recorded thereon, wherein the information, when read by the machine, causes the machine to perform the following steps:

receiving a transcript of a current communication with a customer at a geo-locale;

retrieving a first set of transcripts of historic communications associated with the customer and a second set of transcripts of historic communications occurred at the geo-locale;

computing a prompt input vector based on

a personalized contextual vector obtained based on the transcript and the first set of transcripts, wherein the personalized contextual vector represents personal characteristics of the customer exhibited in the current and the historic communications,

a geo-localized contextual vector obtained based on the second set of transcripts, wherein the geo-localized contextual vector represents characteristics of the geo-locale;

computing a current text vector based on the transcript to characterize the current communication;

generating, by a language model, a prompt based on the prompt input vector and the current text vector, wherein the language model is pretrained via deep learning to predict a next sentence in the current communication based on contexts of the current communication and historic communications captured by the current text vector and the prompt input vector, respectively, and wherein the language model learns, via training, to predict in different contexts;

creating an interactive voice response (IVR) communication based on the prompt for responding to the customer; and

sending the IVR communication to the customer.

8 . The medium of claim 7 , wherein the computing the prompt input vector comprises:

processing the transcript and the first set of transcripts to generate a first processed result;

estimating a first plurality of features characterizing the customer based on the first processed result; and

computing the personalized contextual vector based on the first plurality of features and personalized embeddings previously trained via machine learning.

9 . The medium of claim 8 , wherein the first plurality of features include at least one of:

a first level intent of the customer estimated based on each phrase identified in the first processed result;

a second level intent of the customer estimated based on each session of communication transcript identified in the first processed result;

an emotional state of the customer estimated based on at least one of the first level intent, the second level intent, and the first processed result; and

at least one unanswered question detected from the first processed result.

10 . The medium of claim 8 , wherein the information, once read by the machine, further causes the machine to perform:

processing the second set of transcripts to generate a second processed result;

estimating a second plurality of features characterizing the geo-locale based on the second processed result;

computing the geo-localized contextual vector based on the second plurality of features and geo-localized embeddings previously trained via machine learning;

generating the prompt input vector by combining the personalized contextual vector and the geo-localized contextual vector.

11 . The medium of claim 10 , wherein the second plurality of features include at least one of:

at least one geo-localized intent estimated with respect to the geo-locale based on the second processed result;

at least one intent-based event level detected based on the second processed result, wherein each of the at least one intent-based event level characterizes a corresponding one of the at least one geo-localized intent; and

an emotional state associated with the geo-locale estimated based on the second processed result.

12 . The medium of claim 7 , wherein the creating an interactive voice response (IVR) communication based on the prompt comprises:

generating an IVR communication in a textual form based on the prompt;

converting the IVR communication in the textual form into a voice signal based on text-to-speech transformation; and

outputting the voice signal as the IVR communication.

13 . A system, comprising:

a textual feature vector generator implemented by a processor and configured for

receiving a transcript of a current communication with a customer at a geo-locale,

retrieving a first set of transcripts of historic communications associated with the customer and a second set of transcripts of historic communications occurred at the geo-locale,

computing a prompt input vector based on

a personalized contextual vector obtained based on the transcript and the first set of transcripts, wherein the personalized contextual vector represents personal characteristics of the customer exhibited in the current and the historic communications, and

a geo-localized contextual vector obtained based on the second set of transcripts, wherein the geo-localized contextual vector represents characteristics of the geo-locale, and

computing a current text vector based on the transcript to characterize the current communication;

a dynamic prompt generator implemented by a processor and configured for generating, by a language model, a prompt based on the prompt input vector and the current text vector, wherein the language model is pretrained via deep learning to predict a next sentence in the current communication based on contexts of the current communication and historic communications captured by the current text vector and the prompt input vector, respectively, and wherein the language model learns, via training, to predict in different contexts; and

an IVR communication generator implemented by a processor and configured for

creating an interactive voice response (IVR) communication based on the prompt for responding to the customer, and

sending the IVR communication to the customer.

14 . The system of claim 13 , wherein the computing the prompt input vector comprises:

processing the transcript and the first set of transcripts to generate a first processed result;

estimating a first plurality of features characterizing the customer based on the first processed result; and

computing the personalized contextual vector based on the first plurality of features and personalized embeddings previously trained via machine learning.

15 . The system of claim 14 , wherein the first plurality of features include at least one of:

a first level intent of the customer estimated based on each phrase identified in the first processed result;

a second level intent of the customer estimated based on each session of communication transcript identified in the first processed result;

an emotional state of the customer estimated based on at least one of the first level intent, the second level intent, and the first processed result; and

at least one unanswered question detected from the first processed result.

16 . The system of claim 14 , wherein the computing the prompt input vector further comprising:

processing the second set of transcripts to generate a second processed result;

estimating a second plurality of features characterizing the geo-locale based on the second processed result;

computing the geo-localized contextual vector based on the second plurality of features and geo-localized embeddings previously trained via machine learning;

generating the prompt input vector by combining the personalized contextual vector and the geo-localized contextual vector.

17 . The system of claim 16 , wherein the second plurality of features include at least one of:

at least one geo-localized intent estimated with respect to the geo-locale based on the second processed result;

at least one intent-based event level detected based on the second processed result, wherein each of the at least one intent-based event level characterizes a corresponding one of the at least one geo-localized intent; and

an emotional state associated with the geo-locale estimated based on the second processed result.

18 . The system of claim 13 , wherein the creating an interactive voice response (IVR) communication based on the prompt comprises:

generating an IVR communication in a textual form based on the prompt;

converting the IVR communication in the textual form into a voice signal based on text-to-speech transformation; and

outputting the voice signal as the IVR communication.

Assignments (1)

Continuity (1)

References Cited (14)