Automatic Call Categorization via Regressive Probability Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current call categorization techniques in call centers are unable to provide consistently accurate categorization due to the complexity of call content and high cognitive load on agents, leading to increased after-call workload.
Innovation Solution
An automatic call categorization system that uses a call-to-text processing engine with automatic speech recognition to transcribe calls in real-time, annotate the text with speaker identities, and employ a regressive probability calculation to assign categories based on predefined phrases, displayed in a graphical user interface for summary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual call categorization by agents is used, then call details can be captured accurately, but cognitive load on agents increases and after-call workload increases
Solution Approach 1:
The system enables self-service categorization by automatically analyzing call content and assigning categories without human intervention. The call categorization module processes transcribed call text, identifies key entities and issues, and autonomously determines the appropriate call category, freeing agents from manual categorization tasks while maintaining consistent accuracy.
Solution Approach 2:
The patent replaces the mechanical manual process of agent-based categorization with an automated computational system. The call to text processing engine converts spoken call content into structured text, which is then processed by the call categorization module using predefined rules and algorithms to determine categories, substituting human cognitive effort with automated text analysis.
2Measurement precision
If manual call categorization by agents is used, then call details can be captured accurately, but after call workload for agents increases
Solution Approach 1:
The system performs categorization during or immediately after the call while the content is still fresh in the agent's mind and the call audio is readily available. The call to text processing engine continuously transcribes the call, and the categorization module processes this text in real-time or near real-time, completing the categorization task before the agent needs to log call details, thereby eliminating post-call categorization workload.
3Productivity
If current call categorization techniques are used, then some categorization can be achieved, but consistently accurate categorization cannot be provided due to complexity of call content
Solution Approach 1:
The system segments the complex call content into manageable components by first transcribing the audio into text, then identifying specific entities, issues, and key phrases separately. The call categorization module divides the categorization task into multiple steps: extracting call entities, analyzing issue types, and assigning categories based on predefined rules, making the complex process systematic and consistent.
Solution Approach 2:
The system incorporates feedback mechanisms where the call categorization module continuously refines its categorization based on the analyzed call content and predefined categorization rules. The system can learn from categorization outcomes and adjust its analysis, ensuring consistently accurate categorization even as call content complexity varies.
Data Source
AI summary
A system for categorizing a call between an agent and a caller comprises at least one processor and a memory communicably coupled to the at least one processor. The memory comprises computer executable instructions, which, when executed by the at least one processor implement a method as follows. A call document comprising text of the call between the agent and the caller is received by the system. The system categorizes the call into at least one class using regressive probability analysis of the call document. The system splits the call document to at least two portions, the at least two portions comprising a call header and a call body, and thereafter, using rule-based entity extraction, the system extracts a mandatory entity from the call header and an optional entity from the call body.