Newspaper Article Retrieval Systems for NTCIR2 Workshop

Title: Newspaper Article Retrieval Systems for NTCIR2 Workshop

Research Question: Can two different systems be developed to effectively retrieve newspaper articles based on user queries for various tasks?

Methodology: The researchers developed two systems, System A and System B, to retrieve newspaper articles for different tasks. System A was an enhanced version of the system used for the first NTCIR Workshop's IR tasks and the IREX Workshop's IR tasks. It participated in JJ and CC tasks and achieved high average precisions. System B was a newly developed system with fewer free parameters. It participated in JJ, JE, EE, EJ, and CC tasks and also achieved high average precisions.

Results: System A performed well on the CC tasks due to the characteristics of newspapers, such as the title or the first sentence of the body of an article. System B achieved good results on JJ, EE, JE, and EJ tasks.

Implications: The results suggest that both systems can be effective in retrieving newspaper articles based on user queries for various tasks. The success of System B, which used fewer free parameters, indicates that robust systems can be developed with less manual tuning. The systems' performance on different tasks highlights the importance of understanding the characteristics of the document collection and adjusting the retrieval methods accordingly.

Link to Article: https://arxiv.org/abs/0103013v1 Authors: arXiv ID: 0103013v1