FOREST – Focused object retrieval by exploiting significant tag paths
Introduction
FOREST is a Java system for extracting the main content (the article) of Web pages, developed by Marilena Oita during her PhD. Given a collection of structurally similar pages of the same Web site (typically, pages linked from the site's Web feed), it identifies the tag paths that are significant for the information content of the pages, and uses them to extract the article of each page. It is described in the following publication:
Download
The code and the dataset are provided as is, as they were left in 2014, for research purposes; they were originally distributed from the Web site of the DBWeb team at Télécom ParisTech, which no longer exists.
- FOREST: Java sources (Maven project), third-party libraries, and the reimplementation of the CETR baseline used in the experiments (ZIP)
- RED, the RSS-based Experimental Dataset (ZIP, 17 MB): sample pages of 91 blogs and news Web sites collected through their Web feeds, with, for each site, a set of annotated pages giving the gold standard of the extraction; see the description of the dataset.