Skip to content

Latest commit

 

History

History
15 lines (8 loc) · 2.56 KB

closing.md

File metadata and controls

15 lines (8 loc) · 2.56 KB

Welcome to this final Annif tutorial video. In this video, we will close the tutorial and give some information about what you can do to learn more.

The tutorial consisted of core exercises as well as optional ones. If you haven’t yet completed all the optional exercises, you can still consider completing them. The optional exercises demonstrate some useful techniques such as more powerful machine learning models that can give you better quality results than the simple models covered in the core exercises.

Along the way, you should have found out how well each of the models you’ve trained perform in evaluations. Here you can find the expected scores for each of the models, including the ones covered only in optional exercises. On the left side, you can see the NDCG scores for both data sets and on the right, the F1 scores for the same data sets. As you can see, the shapes of the diagrams are very similar, so both metrics are measuring approximately the same thing.

The interesting part here is that the ensembles in general perform at least as well as the individual models. We can also see that the Omikuji model, which was covered in an optional exercise, gives extremely good results already on its own, and if we include it in an ensemble, the result improves further. The best scores can be achieved with a neural network ensemble that includes all the basic model types.

You should now have completed the hands-on part of the tutorial. If you got stuck on an exercise, want to ask questions or just to discuss the topics of this tutorial, you can consider joining an online interactive session that we arrange from time to time. Please see the Annif-tutorial GitHub repository for details on the schedule and registration.

We have prepared a feedback survey for you to fill in, so that we get to know how to improve the tutorial in the future. Please take a few minutes to fill in the feedback form. The link to the survey is in the Annif-tutorial GitHub repository near the end of the exercise overview page.

Finally, you should also consider joining the annif-users group. It is a web forum and mailing list hosted on Google Groups where you can ask questions, share your experiences and get to know other people who use Annif. If you do something interesting with Annif, please write a message there to let others know about it! The user group is also used for announcements about new Annif releases and other news relevant to the Annif community.

This was the final video in this hands-on tutorial. Thank you for watching and for completing the tutorial. We hope to hear back from you.