News Commentary Generation via Target Cluster Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing news commentary generation methods are inefficient, as they often require processing multiple news clusters for a single event, leading to CPU, storage, and I/O resource wastage, and result in logical confusion due to spliced sentences with semantic differences.
Innovation Solution
A method and apparatus that determine a target news cluster based on relevance and quality, generating a commentary only for the most suitable news within that cluster, reducing resource usage and minimizing logical confusion by selecting the highest-scoring news for commentary generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple news clusters are processed for a single event, then comprehensive coverage of news is improved, but CPU, storage, and I/O resource consumption increases
Solution Approach 1:
The patent extracts and processes only the most relevant news cluster (target news cluster) from multiple clusters for each event, rather than processing all clusters. This extraction is based on determining a target news cluster from multiple clusters, thereby reducing CPU, storage, and I/O resource consumption while maintaining comprehensive coverage through selective processing of the most important cluster.
Solution Approach 2:
The patent applies local quality by differentiating the processing level for different news clusters. The target news cluster receives full processing (commentary generation), while other clusters are either skipped or processed at a reduced level. This selective approach maintains high coverage for the most relevant event while reducing overall resource consumption.
2Reliability
If multiple news clusters are processed for a single event, then comprehensive coverage of news is improved, but storage resource consumption increases
Solution Approach 1:
The patent extracts and stores only the target news cluster rather than all news clusters. By determining a single target news cluster from multiple clusters and processing only that one, the system significantly reduces storage resource consumption while maintaining comprehensive coverage through the selection of the most relevant cluster.
3Reliability
If multiple news clusters are processed for a single event, then comprehensive coverage of news is improved, but I/O resource consumption increases
Solution Approach 1:
The patent extracts and processes only the target news cluster from multiple clusters, thereby reducing I/O resource consumption. The system determines a target news cluster and processes only that one, which significantly reduces input/output operations compared to processing all news clusters while maintaining comprehensive coverage.
4Quantity of substance
If sentences from multiple news clusters are spliced to generate commentary, then comprehensive information is improved, but logical confusion and semantic differences increase
Solution Approach 1:
The patent extracts and processes only the target news cluster rather than splicing from multiple clusters. By selecting a single target cluster and generating commentary from that one, the system avoids logical confusion and semantic differences that arise from splicing multiple sources, while still providing comprehensive information through the selected cluster.
Solution Approach 2:
The patent segments the news processing into distinct steps: determining the target news cluster, then generating commentary from that single cluster. This segmentation prevents the logical confusion that would result from mixing multiple clusters, as the commentary generation is performed on a single, coherent source.
Data Source
AI summary
Embodiments of the present disclosure provide a method and apparatus for generating a commentary. The method may include: acquiring at least one news cluster composed of pieces of news generated within a first preset time length, the pieces of news in the news cluster direct to a given news event; determining a target news cluster based on the at least one news cluster; determining, for each piece of news in the target news cluster, a score of being suitable for generating a commentary for the piece of news; and generating, based on a piece of target news, a commentary for the target news cluster, where the piece of target news is a piece of news having a highest score of being suitable for generating a commentary in the target news cluster.