Real-Time Contact Center Call Summarization Using Intent Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact center agents face challenges in creating accurate and efficient summaries of calls due to the time-consuming nature of traditional transcription methods and the inaccuracy of speech recognition, leading to incomplete or varied summaries that are difficult to comprehend and analyze.
Innovation Solution
A system that includes a speech-to-text generating unit, an intent recognizing unit using machine learning, and an agent user interface for real-time transcription, intent matching, and summary editing, allowing agents to confirm, edit, and refine call summaries, ensuring accuracy and uniformity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional transcription methods are used to create call summaries, then the summaries can be generated, but the process is time-consuming and reduces agent productivity
Solution Approach 1:
The system performs preliminary transcription and intent recognition during the call itself, so that the summary is already prepared or nearly prepared by the time the call ends. The speech-to-text generating unit transcribes the call in real-time, and the intent recognizing unit identifies key information as it occurs, eliminating the need for post-call summary creation.
Solution Approach 2:
The manual mechanical process of reading and writing summaries is replaced with automated AI systems. The speech-to-text generating unit converts speech to text automatically, and the intent recognizing unit uses machine learning to identify and structure key information, replacing the agent's manual effort with intelligent automation.
2Reliability
If agents manually create call summaries, then summaries can be produced, but accuracy and completeness suffer due to agent workload pressure
Solution Approach 1:
The manual process of creating accurate summaries is replaced with automated AI systems that can process and analyze call content without fatigue or distraction. The intent recognizing unit uses machine learning to accurately identify key information, entities, and customer intent, ensuring consistent accuracy regardless of agent workload.
Solution Approach 2:
The system provides feedback to agents through the agent user interface, allowing them to review, confirm, or correct the automatically generated summary. This feedback loop ensures accuracy while maintaining high productivity, as agents only need to verify rather than create summaries from scratch.
3Measurement precision
If traditional text summarization techniques are used on call transcripts, then summaries can be generated, but the unstructured nature of human conversation and speech recognition errors make these techniques ineffective
Solution Approach 1:
Traditional text summarization techniques are replaced with speech-to-text generation and intent recognition systems specifically designed for conversational data. The speech-to-text generating unit is optimized for transcribing human speech, and the intent recognizing unit uses machine learning models trained on conversational patterns, making the system specifically suited for call summarization rather than generic text processing.
Solution Approach 2:
The system changes the approach from text-based summarization to speech-based intent recognition. Instead of analyzing transcribed text for summary generation, the system directly identifies customer intent, entities, and key information from the speech patterns and conversational structure, adapting the processing method to the unstructured nature of human dialogue.
4Loss of information
If complete call transcripts are stored in databases, then all information is preserved, but the length and difficulty of comprehension reduce ease of operation
Solution Approach 1:
The system extracts only the essential information from complete call transcripts, identifying key entities, customer intent, and important details while leaving out unnecessary conversational filler. This extraction process preserves critical information while eliminating redundancy, making the summary both complete in essential aspects and easy to read.
Solution Approach 2:
The call summary is segmented into structured components including customer intent, extracted entities, and key information points. This segmentation organizes the information in a logical, easy-to-navigate format that maintains completeness while improving readability and accessibility for agents and analytics systems.
Data Source
AI summary
A method for creating a textual summary of a call includes transcribing speech to text in real time using a speech-to-text generating unit configured for execution upon one or more data processors, automatically matching, in real-time, text to predetermined intents and extracted entities using an intent recognizing unit for execution upon the one or more data processors, automatically mapping the predetermined intents and extracted entities into a call summary using one or more mapping functions, and displaying the call summary using an agent user interface for execution upon the one or more data processors. A contact center call summarization system may include a contact center communication device, a speech-to-text generating unit, an intent recognizing unit, and an agent user interface.