WEBIS-WEBSEG-20 ALGORITHM SEGMENTATIONS
=======================================

Dataset of segmentations and evaluation results on the Webis-WebSeg-20 dataset.

  - Check for up-to-date resources: https://webis.de/data.html#webis-webseg-20-algorithm-segmentations
  - The employed Webis-WebSeg-20 dataset: https://webis.de/data.html#webis-webseg-20 / https://doi.org/10.5281/zenodo.3354902
  - WARC files for the web pages in the Webis-WebSeg-20 dataset are contained in the webis-web-archive-17: https://webis.de/data.html#webis-web-archive-17 / https://doi.org/10.5281/zenodo.1002203


Contents
--------
All ZIP archives extract to a common directory hierarchy of `webis-webseg-20/<pageid>` that blends with the structure of the Webis-WebSeg-20. For convenience, we split the resources into different parts.

  - webis-webseg-20-algorithm-segmentations.zip
    Contains the segmentation of each algorithm. Uses the file format of original Webis-WebSeg-20, see https://github.com/webis-de/cikm20-web-page-segmentation-revisited-evaluation-framework-and-dataset
  - webis-webseg-20-algorithm-evaluations.zip
    Contains two files for each page:
      - evaluation.csv
        Lists the achieved BCubed Precision, Recall, and F1 for each combination of algorithm and atomic element type.
      - num-segments.csv
        Lists the number of segments for each algorithm
  - webis-webseg-20-meier-models.zip
    Contains the models trained for the approach of Meier et al., see https://github.com/webis-de/ecir21-an-empirical-comparison-of-web-page-segmentation-algorithms#meier-et-al
  - webis-webseg-20-4096px.zip
    Contains the intermediate data created for the approach of Meier et al., see https://github.com/webis-de/ecir21-an-empirical-comparison-of-web-page-segmentation-algorithms#meier-et-al

In addition, the dataset contains the following files:
  - evaluation.csv
    The averaged values from the evaluation.csv files in webis-webseg-20-algorithm-evaluations.zip, as used in the plots of the paper.
  - num-segments.csv
    The averaged values from the num-segments.csv files in webis-webseg-20-algorithm-evaluations.zip, as used in the plots of the paper.
  - webis-webseg-20-folds.txt
    Contains for each cross-validation fold (0 through 9) the contained page/task IDs (only used for the approach of Meier et al.)


Authors
-------
 - Johannes Kiesel,  johannes.kiesel@uni-weimar.de
 - Lars Meyer,       lars.meyer@uni-weimar.de
 - Florian Kneist,   fkneist@gmail.com
 - Benno Stein,      benno.stein@uni-weimar.de
 - Martin Potthast,  martin.potthast@uni-leipzig.de