Automated recognition system for natural language understanding

• US 10,147,419 B2
• Filed: 08/30/2016
• Issued: 12/04/2018
• Est. Priority Date: 01/05/2011
• Status: Active Grant

• Current Assignee: Interactions, LLC
• Original Assignee: Interactions, LLC
• Inventors: Yeracaris, Yoryos, Lapshina, Larissa, Carus, Alwin B.
• Primary Examiner(s): Han, Qi
• Application Number:
  US15/251,868
• Publication Number:
  US 20160372109A1
• Time in Patent Office: 826 Days
• Field of Search: 704257, 704231, 704235, 704244, 704250, 704255
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 2015/0635 updating or merging of old and new templates; Mean values; Weighting
• CPC Class Codes: G10L 2015/0638 Interactive procedures
• CPC Class Codes: H04M 2203/2061 Language aspects
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00
• CPC Class Codes: H04M 3/5166 in combination with interactive voice response systems or voice portals, e.g. as front-ends

Abstract

An interactive response system directs input to a software-based router, which is able to intelligently respond to the input by drawing on a combination of human agents, advanced recognition and expert systems. The system utilizes human “intent analysts” for purposes of interpreting customer input. Automated recognition subsystems are trained by coupling customer input with IA-selected intent corresponding to the input, using model-updating subsystems to develop the training information for the automated recognition subsystems.

Claims

1. A system for processing an interaction with a person, comprising:

a routing processor configured to receive data representing an input provided by the person;

an analyst user interface device in communication with the routing processor, configured to present to at least one human analyst information related to the input in perceptible form and to accept an intent from the at least one human analyst; and

a training subsystem configured to;

determine a level of end-user traffic;

based on the level of end-user traffic, determine a number of human intent analysts to assign to interpret user input;

receive the information and the intent;

select a target vocabulary based on a business meaning corresponding to the input as determined by at least one of the human intent analysts;

train a first model used by a training ASR and generated based at least in part on the information and the intent and using the target vocabulary to interpret the input, the training ASR configured to generate statistics responsive to the information; and

train a second model used by the real-time ASR responsive to the statistics.

2. The system of claim 1, the system further comprising a real-time automated speech recognizer (ASR) in communication with the routing processor and configured to receive therefore the data, wherein the training subsystem is configured to train the real-time ASR by updating the second model.

3. The system of claim 1, the system further comprising a real-time automated speech recognizer (ASR) in communication with the routing processor and configured to receive therefore the data, wherein the training subsystem is configured to continue training the real-time ASR responsive to the real-time ASR not reaching a performance threshold.

4. A computer-implemented method for operating an interactive response system comprising:

determining a level of end-user traffic;

based on the level of end-user traffic, determining a number of intent analysts to assign to interpret user input

receiving data representing an input from a person;

using a processor, automatically presenting information relating to the input to at least one of the intent analysts through an analyst user interface;

accepting an intent from the at least one intent analyst through the analyst user interface;

providing the information and the intent to a training subsystem;

accepting from the training subsystem a target vocabulary selected based on a business meaning corresponding to the input as determined by the at least one intent analyst;

accepting from the training subsystem a training model used by a training automated speech recognizer (ASR), the training ASR generated responsive to the information and the intent and using the target vocabulary to interpret the input;

accepting from the training ASR statistics generated responsive to the information; and

training, via the statistics, a second model used by a real-time ASR in order to improve performance thereof.

5. The computer-implemented method of claim 4, wherein the training comprises updating the second model.

6. The computer-implemented method of claim 4, wherein the training comprises testing performance of the real-time ASR and continuing training responsive to the performance not exceeding a performance threshold.

7. A system for managing interactions with a person, comprising non-transitory computer storage media storing programming instructions executable by at least one processor for:

determining a level of end-user traffic;

based on the level of end-user traffic, determining a number of intent analysts to assign to interpret user input

receiving data representing an input from a person;

using a processor, automatically presenting information relating to the input to at least one of the intent analysts through an analyst user interface;

accepting an intent from the at least one intent analyst through the analyst user interface;

providing the information and the intent to a training subsystem;

accepting from the training subsystem a target vocabulary based on a business meaning corresponding to the input as determined by the at least one intent analyst;

accepting from the training subsystem a training model used by a training automated speech recognizer (ASR), the training ASR generated responsive to the information and the intent and using the target vocabulary to interpret the input;

accepting from the training ASR statistics generated responsive to the information; and

training, via the statistics, a second model used by a real-time ASR in order to improve performance thereof.

8. The system of claim 7, wherein the training comprises updating the second model.

9. The system of claim 7, wherein the training comprises testing performance of the real-time ASR and continuing training responsive to the performance not exceeding a performance threshold.