Fine-grained topic detection in news search results
Proceedings of the 27th Annual ACM Symposium on Applied Computing
| Hi-index | 0.00 |
In the era of web 2.0, everyone can create and update content, and everyone can host a personal web site with little effort, making it hard to gather valuable information from different web sites. With RSS, people can read information from different resources in a uniform way, and in a single tool, such as an RSS reader. However, most of the RSS readers only display items in chronological order, which doesn't work well when users are inundated with too many items in the feeds. We propose using recommendation to help people find items in an RSS reader. Specifically, we consider profiled based features (i.e., text similarity and favorite fraction), update frequency, as well as Post Rank values for RSS recommendation. Experimental results indicate that favorite fraction and update frequency perform better than text similarity. Additionally, we also study the effect of feature combination and find that the combination of similarity and favorite fraction performs the best.