Time-Series Query Focused Summarization for Customer Service Tickets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing large amounts of data associated with customer service requests in wireless communication services is challenging, as existing summarization methods, such as computer-generated summaries and Query Focused Summarization (QFS) models, often produce inaccurate or outdated summaries due to the chronological nature of ticket documents.
Innovation Solution
The implementation of a Time-Series Query Focused Abstractive Summarization (TQFS) model, which computes document relevancy scores based on a combination of query similarity and timestamp metrics, ranks documents accordingly, and inputs them into a document summarizer to generate a summary that balances context and recent information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing summarization methods are used to process chronological ticket documents, then the summarization process can be performed, but the generated summaries are inaccurate or outdated
Solution Approach 1:
The patent applies dynamics by making the document selection process adaptive rather than static. The system dynamically adjusts which documents to include in summarization based on their temporal characteristics and relevance to the customer inquiry, allowing the summary to remain current and accurate as new information becomes available
Solution Approach 2:
The patent changes the parameter of document selection from simple chronological ordering to a relevance-based ranking system. By transforming the selection criterion from time-sequential to relevance-weighted, the system produces summaries that are both timely and accurate, addressing the contradiction between these two requirements
2Loss of information
If all chronological documents are processed for summarization, then comprehensive context is captured, but the summary includes outdated information and becomes less readable
Solution Approach 1:
The patent extracts only the most relevant documents from the complete chronological set based on their relevance to the customer inquiry. By selecting and extracting key documents rather than processing all documents equally, the system maintains context completeness while improving readability by excluding outdated or less relevant information
Solution Approach 2:
The patent applies local quality by treating different documents differently based on their individual relevance characteristics. Instead of uniform processing, each document is evaluated and weighted according to its specific relevance to the inquiry, allowing the summary to focus on locally important information while maintaining overall context
3Measurement precision
If Query Focused Summarization models are used to improve summary relevance, then query-related information is enhanced, but the chronological nature of ticket documents causes outdated information to be included
Solution Approach 1:
The patent combines QFS with dynamic document selection based on temporal relevance. The system dynamically weights documents not only by query similarity but also by their temporal characteristics, ensuring that query-relevant information is prioritized while maintaining information currency by giving appropriate weight to recent documents
Solution Approach 2:
The patent creates a composite scoring mechanism that combines query relevance metrics with temporal metrics. This composite approach integrates multiple dimensions (query matching and time recency) into a unified document selection criterion, producing summaries that are both query-relevant and current
Data Source
AI summary
A device may include a processor configured to obtain a time series of documents; determine query scores for particular documents of the time series of documents based on a set of query terms; and determine time scores for the particular documents based on timestamps associated with the particular documents. The processor may be further configured to compute document relevancy scores for the particular documents based on a combination of the query scores and the time scores for the particular documents; order the particular documents based on the computed relevancy scores; and generate a document summary by applying a document summarizer applied to the ordered particular documents.