search

webis-de / ecir21-an-empirical-comparison-of-web-page-segmentation-algorithms

GNU Lesser General Public License v2.1
26 stars 6 forks source link

evaluation #6

Open Bill-Ren opened 1 year ago

Bill-Ren commented 1 year ago

Thanks for a great job. Your paper uses the Webis-WebSeg-20 dataset, which contains 8490 web pages, for the empirical comparison of web page segmentation algorithms. I would like to ask if you divide the training set and the test set when evaluating the model? Or do you test the entire data set during the test, and then get the indicators in the paper?

johanneskiesel commented 1 year ago

Hi,

Sorry for not yet answering you in the other issue. It needs a bit more time from my side, and I could not find that yet.

Regarding your question here:

The only model we trained is that of Meier et al. The other ones we used out-of-the-box. For Meier et al. we use a 10 fold cross validation setup: https://github.com/webis-de/ecir21-an-empirical-comparison-of-web-page-segmentation-algorithms/tree/master/algorithms/meier#train-the-model . If you want to use the same folds, you can find the fold assignment here: https://zenodo.org/record/4337578/files/webis-webseg-20-folds.txt

I hope this answers your question.

Sorry again for being so slow for the other one!

Bill-Ren commented 1 year ago

OK, thank you for your patient answer. Because I have been studying this data set recently, I asked a lot of questions. Thank you again for your patient answer.