I was trying to watch the Icelandic show Katla, and I noticed that non-English letters seem to be treated as non-word characters for purposes of segmentation. In this screenshot, you can see that it treats the i as a word, despite the fact that it’s just in between two ðs.
I have not found any non-English characters that are correctly recognised.
Given that non-English chacacters work fine in the Norwegian show I watched (Ragnarok), my hypothesis is that each language’s word segmentation rules are configured separately, and Icelandic is using a default English-Latin segmentation.