dbpedia / DBTax

DBTax project
Apache License 2.0
3 stars 4 forks source link

Threshold Calculations #10

Open sm86 opened 7 years ago

sm86 commented 7 years ago

I have used this approach for calculating threshold. In given doc, I have checked if slope<4 but when I have seen the Wiki-en database, results have showed slope<50 as ideal. Declaring a cutoff slope value will ideally depend on total no. of elements in database too. For embedded H2 database, we would have a different cutoff slope. Any suggestions on how to go about implementing this.

marfox commented 7 years ago

In general, I think your approach makes sense, but at the end it boils down to manually setting 2 values, instead of 1 in the original approach: 1 for the slope and 1 for what you called the count. You still have to plot the distribution and decide based on visual results.

While automating the optimal threshold calculation would be a great step forward, I recognize it is quite challenging and may still be out of scope for this project. I suggest to manually find one value that should be better than the original one

sm86 commented 7 years ago

Thanks @marfox . Sure, in that case I will generate a csv file, which can be used to plot.
Would you like me to include a graph too, I can use JSP to add that part to project.

marfox commented 7 years ago

Sounds good

Il 06 lug 2017 8:19 AM, "Shashank Motepalli" notifications@github.com ha scritto:

Thanks @marfox https://github.com/marfox . Sure, in that case I will generate a csv file, which can be used to plot. Would you like me to include a graph too, I can use JSP to add that part to project.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/dbpedia/DBTax/issues/10#issuecomment-313306386, or mute the thread https://github.com/notifications/unsubscribe-auth/AEnTqEE7SeOk3LCcRnJGoiW8glNEUxJuks5sLHxpgaJpZM4OOBDF .

sm86 commented 7 years ago

Would you like this JSP app running on Tomcat integrated in same project or a different project?