Conversation Summary Generation for Customer Service Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Customer service agents face difficulties in maintaining clear records of multiple simultaneous chat consultations, leading to challenges in smoothly managing conversations and providing effective service.
Innovation Solution
An electronic device generates summary information for conversation texts using pre-trained AI models to extract and summarize key entities or topics, adjusting scores based on accuracy and contextual relevance, allowing agents to quickly grasp conversation content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single customer service agent chats with a large number of customers at the same time, then productivity is improved, but the agent cannot clearly remember the contents of the consultation and it becomes difficult to proceed smoothly
Solution Approach 1:
The system performs preliminary summarization of conversation texts in real-time as conversations progress, creating condensed versions of customer interactions before the agent needs to review them. This preliminary action allows the agent to quickly grasp essential information without having to remember all details of multiple simultaneous conversations.
Solution Approach 2:
The electronic device acts as an intermediary between the customer service agent and the conversation texts. It automatically generates and provides summary information to the agent, serving as a mediator that bridges the gap between the large volume of conversation data and the agent's cognitive processing capabilities.
2Reliability
If the agent tries to remember all conversation details, then accuracy of service is improved, but it becomes difficult to manage multiple chats efficiently
Solution Approach 1:
The system extracts only the essential and relevant information from complete conversation texts, creating condensed summaries that contain key details needed for accurate service delivery. This extraction process removes unnecessary information while preserving the essential elements required for reliable customer service.
Solution Approach 2:
The system transforms the parameter of information representation from detailed complete conversations to condensed summary forms. By changing the parameter of how information is presented (from full text to summarized key points), the agent can maintain service accuracy while managing multiple chats more efficiently.
3Measurement precision
If the agent reviews complete conversation texts, then understanding of customer inquiries is improved, but time consumption increases and response speed decreases
Solution Approach 1:
Instead of requiring the agent to review complete conversation texts, the system provides partial information in the form of condensed summaries that contain the essential understanding needed for customer inquiries. This partial action approach maintains sufficient understanding while dramatically reducing the time required to review conversation content.
Solution Approach 2:
The system performs the summarization and information extraction in advance, before the agent needs to review the conversation content. This preliminary processing of conversation texts into summaries allows the agent to quickly access understood information without spending time on detailed review.
Data Source
AI summary
A method, performed by an electronic device, of generating summary information of a conversation text, includes: obtaining at least one first summary text extracted from at least one conversation text of a previous conversation; extracting at least one second summary text from a conversation text of a current conversation; identifying, from among the at least one first summary text, a summary text having a same type as a type of the at least one second summary text; adjusting a score indicating an accuracy of the identified summary text; selecting, from among the at least one first summary text and the at least one second summary text, a summary text corresponding to each of at least one specified type based on a score of each summary text; and generating the summary information based on the selected summary text.