.png){kind=icon}
.png){kind=icon}
Return
Requirements Analysis for Sentence Feature-Based Multi-document Update Summarization
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-032-09037-9_23.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Automatic Text Summarization aims to generate a short summary that covers all aspects, is not redundant, and shows the most relevant information from a Single Document Text Summarization (SDTS) or Multi-Document Text Summarization (MDTS). MDTS is a more challenging task than SDTS because input documents are likely to contain more contradictory, redundant, and complementary information. This work proposes a method for generating MDTS based on extracting text features to address the requirements of coverage, reduction of redundancy, and relevance of sentences. Moreover, we test on the TAC08 dataset in order to know the performance for capturing the new information in update summaries of 100 words, where readers are considered to have the basic information about the news and require only recent updates regarding that news. We evaluated the proposed method with the ROUGE system. Our results are compared with the heuristics and participants in the TAC08 workshop, recent methods, and Large Language Models, achieving competitive results. In addition, we determine the importance of each requirement in the update MDTS, in order to establish a priority hierarchy between them. This result could be a benchmark for assigning importance to requirements in future research.
Keywords:
Coverage
Reduction of redundance
Relevance
Update Summaries
Multi-Document Text Summarization
Journal
A
IF:
0
Papers:
31
Citations:
0