.png){kind=icon}
.png){kind=icon}
Return
A Multi-Document Summarization Method for Customer Feedback Based on Large Language Models
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-031-93251-9_15.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the era of big data, automated summarization helps users quickly extract key information from large amounts of data. Customer feedback is important textual data for companies to improve their products and services. When customer feedback is numerous and diverse, extracting critical information from them can be a time-consuming task, which requires multi-document summarization (MDS) service. However, existing MDS methods always simply enumerate the most critical issues in each feedback, rather than summarizing the core issues of most customers. This results in service personnel being unable to quickly find out the reasons and accurately make corresponding countermeasures. To overcome this challenge, this paper presents a customer feedback summarization method based on large language models (LLMs). We propose a two-stage Event-Triplet-Tree (ETT) prompt: 1) Event triplet extraction, which extracts a list of event triplets from multiple customer feedback, 2) Iterative merging of event triplets, which iteratively grouping, merging, and summarizing triplets in the list until only one refined triplet remains. ETT prompt can provide service personnel with a comprehensive and in-depth overview of customer feedback, allowing them to identify core issues and expectations faster, thereby improving customer satisfaction. Experimental results show that our proposed approach outperforms the general prompt across all datasets.
Keywords:
Generative AI
Large Language Models
Multi-Document Summarization
Customer Feedback
Prompt Engineering
Journal
C
IF:
0
Papers:
18
Citations:
0