Hello everyone,
just tried uploading my lecture slides in .pdf format to AnkiHub AI (German Ankizin deck), but it’s giving me 400 mostly off-topic cards, even when I upload just a single slide.
Has anyone else experienced this issue and knows how to resolve this?
Thanks in advance for your help!
Thank you for reaching out! Our moderators have been notified and will assist you shortly. In the meantime, you can get an instant response from me simply by @ mentioning my username`. Alternatively, start a private conversation with me (see screenshot below). If you ask me a question, I will do my best to respond using only verified information found in the official AnkiHub docs and the Anki manual.
You can also find answers to common questions in our FAQ and Docs.
Hello,
Can you DM the PDF so we can troubleshoot?
Also, I’m assuming the PDF you are uploading has relevant content to the cards. We have found a bug recently where smartsearch results sometimes brings up irrelevant results and we’re in the process of addressing it
Does this happen to any PDF you upload?
Hello,
of course, I sent you the PDF via DM.
Yes, I selected the Ankizin deck and uploaded the slides of thoracic surgery, which have their own section in the deck. This also happens when I only paste the text of one slide into the text search field.
When you paste the text of one slide into the text search field, how many cards do you get? and all of them are irrelevant correct?
Almost every time, I get 400 cards (even with just one slide). Also, when I try to filter them by importance, about 350 of the cards are marked as 100% relevant, even though they sometimes don’t contain a single word from the text prompt.
This is very helpful, thank you for letting us know!
You’re welcome! Is there any way I could fix it?
This is something we would have to fix on our end, we’re planning on addressing this ASAP
Do you know when it’s approximately planned to be addressed?
I don’t believe we’ve been given a time estimate by the devs just yet. I know they’re working on it as we speak
I’ve noticed this too. When it was in Beta, it felt very accurate. However, recently, it has been pulling up so many irrelevant cards—even with specificity set to 100%—that I’ve mostly stopped using it.
If you ever need power points to help trouble shoot this, feel free to DM me!
Can you elaborate further during the beta testing period. How accurate was it? When you uploaded a PDF, how many cards were irrelevant and how many cards were relevant (percentage wise)
Would you say that it randomly got worse?
Were you satisfied with the results during beta testing and now it seems like it’s completely changed?
We have been investigating this and your input would be amazing to us!
During the beta phase, I found that about 90% of the cards were relevant, possibly even more. However, after that—either towards the end of the beta or after release—the relevance of the cards suddenly dropped and has remained consistently worse. Now, I feel that only about 50% of the cards are relevant, and there are cards that should be found but aren’t. The change in performance was abrupt; I remember one week it worked really well, and then the next week, it was suddenly much worse. I can not remember exactly when this change occurred.
One thing I have noticed, that could be irrelevant to the problem but I found interesting, is that if you change the search from only including Step 1 cards to including both Step 1 and Step 2 cards, occasionally, the number of cards it finds is actually lower after the change.
