Automated Call Categorization Using Speech-to-Text and Machine Learning

Solution Overview

Problem

Existing systems face challenges in objectively analyzing and categorizing large volumes of phone calls due to human subjectivity and inefficiency, leading to inaccurate and time-consuming processes for determining customer intent and call outcomes.

Innovation Solution

A system and method for automatically categorizing calls using a call tracking system integrated with a call categorization system, which includes a call waveform processor, transcriber, and analytics processor, employing machine learning models to analyze structured and unstructured data, and allowing for customization through a user interface for building and optimizing call categorization models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If employees manually monitor and listen to calls to gather information about call purpose, then human judgment can be applied to understand customer intent, but the process becomes excessively time-consuming and cannot handle large volumes of calls efficiently

Engineering Contradiction:

Improvecall categorization accuracyVSAvoidcall processing volume

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces the mechanical system of human employees manually listening to and categorizing calls with an automated computer-based system. The system uses speech-to-text conversion to transform audio calls into text, then applies natural language processing and machine learning algorithms to automatically categorize calls by purpose, outcome, and sentiment. This substitution enables the system to process large volumes of calls efficiently while maintaining consistent categorization standards without human subjectivity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The automated call analysis system performs self-service by independently analyzing call recordings, converting speech to text, extracting key information, and categorizing calls without requiring human intervention. The system uses trained machine learning models to automatically determine call purposes, outcomes, and sentiments, enabling businesses to process and analyze calls autonomously at scale.

Inventive Principle:

Principle #25Self-service

2Ease of operation

If employees manually analyze calls to determine customer intent and call outcomes, then subjective human judgment is applied, but this leads to inconsistent and unreliable categorization across different callers

Engineering Contradiction:

Improvecategorization process simplicityVSAvoidcategorization consistency

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent replaces the mechanical system of human employees manually listening to and categorizing calls with an automated computer-based system. The system uses speech-to-text conversion to transform audio calls into text, then applies natural language processing and machine learning algorithms to automatically categorize calls by purpose, outcome, and sentiment. This substitution enables the system to process large volumes of calls efficiently while maintaining consistent categorization standards without human subjectivity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system incorporates feedback mechanisms where categorization results are continuously analyzed and used to refine and retrain the machine learning models. This feedback loop enables the system to learn from past categorizations, improve accuracy over time, and maintain consistent standards across all calls while adapting to new patterns and nuances in customer communications.

Inventive Principle:

Principle #23Feedback

3Productivity

If a simple automated keyword scanning system is used to categorize calls, then processing speed increases, but the system lacks accuracy and reliability in determining actual call outcomes

Engineering Contradiction:

Improvecall processing speedVSAvoidcall categorization accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces the mechanical system of human employees manually listening to and categorizing calls with an automated computer-based system. The system uses speech-to-text conversion to transform audio calls into text, then applies natural language processing and machine learning algorithms to automatically categorize calls by purpose, outcome, and sentiment. This substitution enables the system to process large volumes of calls efficiently while maintaining consistent categorization standards without human subjectivity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system transforms the analysis from simple keyword matching to a multi-parameter analysis approach. Instead of relying on single keywords, the system analyzes multiple parameters including speech-to-text conversion accuracy, contextual understanding, sentiment analysis scores, and pattern recognition metrics. This multi-parameter approach enables both high processing speed and accurate categorization by evaluating calls from multiple dimensions simultaneously.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10923127B2System, method, and computer program product for automatically analyzing and categorizing phone calls

Publication Date: 2021.02.16 INVOCA INC

AI summary

Disclosed embodiments relate generally to systems and methods for automatically analyzing and categorizing phone calls for subsequent processing by telecommunications systems. Waveform and/or data analysis may automatically be performed on incoming call records in real or near real time.