You’re right, it’s not just the subtitle grouping. Even with perfectly formed subtitles, the software can tag words wrong. We tag words into these categories using neural networks: https://universaldependencies.org/u/pos/index.html . Tagging is 90-98% accurate, depending on the language, this is ‘state of the art’ at the current time. How we combine these tags is a question, saving a unique item for every tag is probably too much. Actually I forget the exact logic used, but me and Og are about to figure out how to do highlighting for word frequency, so we’ll discuss that a bit now. 
1 Like
We only set 2 sizes at the moment - we use the bigger one for some Asian languages and Arabic. I think Arabic would be a bit hard to read if we used the smaller size. Not sure though. Maybe if we adjusted Arabic font size a little, and made keyboard shortcuts (+/-) to change font size, this would be sufficient?
Do you still have the problem with some items being cut off for Arabic, could you check please?
This is how ‘normal’ size looks for me (hiding half the screen), and yes it’s still getting cut in the side panel.
I switch languages often, so it won’t help if the change is global and the relative difference between scripts is big. This setting should be able to be adjusted independently per script.
Just adjusting the default values won’t help either since what is too big for me would probably be too small for another user.
OK, thanks. What is the video url for the girl playing guitar?
Hi! Um… I was wondering support would be possible for Yandex Translate. It provides transliterations for my target language (which is Hebrew), and I think it does for other languages too (it does for Japanese and Chinese, but not for Greek and Russian from what I’ve seen). It looks like this:
The translations themselves are a little weird, and the transliterations are sometimes erroneous, but it’s up to the reader to discern right from wrong. 
Can you install this extension: https://chrome.google.com/webstore/detail/whatfont/jabopobgcpjmedljpbcaablpmlmfcogm?hl=en
After installing, click on the extension button, and click on the Arabic text in the side panel. It should show something like this:
Let me know what font family (e.g. Roboto, Arial, sans-serif) and size (e.g. 19px) it shows for you.
Hey, I just realized that I had the Chrome extension Wudooh active, which its purpose is enlarging the Arabic font on websites. I completely forgot about its presence (i installed it a long time ago). 
I deactivated it now, and I think both issues are resolved. Sorry for the inconvenience and for wasting your time on that. I’ll let you know if any other issue persists after further testing.
You can mark these issues resolved for now. Sorry, again.
Edit: Without that extension active, the font size in the side panel is a bit too small. That extension made it much more legible (Arabic script is a bit difficult to read when the font size is too small, which is why I needed that extension in the first place).
Link to the extension: https://chrome.google.com/webstore/detail/wudooh-وضوح/nigfaloeeeakmmgndbdcijjegolpjfhn
Anyway, nothing too crucial.
Cool Would be good to make it work with that extension too, but I use some special scrolling mechanism for side panel, so it’s not an easy fix.
I decreased main subs’ font size just slightly, and increased side panel font size for Arabic/Asian langs for the next version. There might be other small issues - font sizes are a bit tricky to get right, so if you notice anything else, let me know ofc.
Yup. But transcription exercises are incredibly powerful to improve your listening abilities, as they force you to decode every single sound.
1 Like
From the point of about 8,000 to 10,000 entries and upwards, the buttons for exporting as JSON and exporting as CVS, no longer work. It’s processing the request for a couple of seconds and then you get the following error:
Maybe it timeouts or something. No file is downloaded.
I’m using those buttons only in order to count my entries in the database, at the moment, so it would be great if I could just have that info displayed instead. It’s just a waste of bandwidth, if I need to check it like that.
I’ll suggest also adding a counter for the number of words you add per day and what you average over the month/year.
If you want to gamifiy a bit your plugin, you can also add the ability to set custom goals for the number of items added per day, and maybe streaks if you want to copy that from Duolingo.
Those features might seem superflues, and just adding extra fat, but you should realize that encouraging motivation is a key aspect of learning a language, if not the most important aspect.
I’ll suggest modifying the behavior of the arrow keys, so that if, for example, you press the backward arrow key, it will only skip to the previous subtitle, if the distance is less than 5 seconds away, but if the distance is longer than that, it should act as skipping 3 seconds backwards/forwards instead.
So a combo behavior, which is more intelligent. Preventing you from jumping great distances (forwards/backwards) unintended and thus easing navigation. That would fit much more closely to the original intent of the user.
You can also add this as a selectable behavior option.
Suggestion for improving guesswork of original audio language, when auto-subtitling is not available.
If no auto-subtitling track is available, and the video includes 2 subtitle tracks, which one of them is English, you can improve your guessing of the appropriate default track to select, by assuming that if auto-subtitling isn’t offered for the video, that the video language is most probably non-English and the other track should be selected by default.
You can further improve your algorithm guesswork by cross-checking, if the non-English subtitle track is among the list of languages which Youtube doesn’t offer speech recognition for, and also by checking the country of origen of that Youtube account. The combinations of these factors can give you a much higher accuracy rates.
You can also use the video title to discern the language from that, if that is better.
It’s enough to just determine that it’s a non-English title (English title doesn’t imply that the video is in English, however).
Case example:
Video in Tagalog with 2 subtitle tracks (1 in English and 1 in Tagalog), and auto-subtitling is not offered.
In this case you would know that Tagalog isn’t supported by speech recognition on Youtube, so you assume Tagalog is the default language. You can further confirm by checking if the location of the account is in the Philippines and by analyzing the video title language, if that helps.
I know it’s in the pipeline, but I just want to add a +1 for “human” translations. There are multiple channels that I subscribe to which have (human-made) subs for both my native language and my target, so I’d love to be able to use them both.
Keep up the good work!
Out of interest, what went into ranking the YouTube channels in the catalogue? It’s a good list, but seems not very good for people using LLwYT given these channels are more for teaching people Japanese instead of watching in Japanese for fun.
My channel on your list for instance has lots of videos from Japanese dramas, anime, and audiobooks along with the text, but it’s not exactly a good channel to watch for fun and I’m interrupting the narrative to look up meanings of Japanese words/phrases.
If you want to find Youtube videos in another language you must search for them in that language. This requires one to know enough in the target language to do that but, hey, that’s part of the fun.
Hello, first off, both of your apps are fantastic. Probably the best language learning apps for intermediate+ language learners. I have a pro account which I use on my Netflix extension, but logging into the same account for YouTube I’m not able to save any items. I’m looking for the star icon like on LLN but am not seeing it. I can see that I’m successfully logged in though because my saved items from LLN are there. Do I need a separte account?
The extension treats multiline subtitles as separate phrases. Example video: https://www.youtube.com/watch?v=eDfMDkgheQY
These 2 lines should be seen on the screen at the same time, but instead they are treated as separate by the extension. They also make you unpause twice if you have Auto-Pause enabled.
Any chance there will be an update anytime soon to add new features (such as marking entire sentences or highlighting specific phrases)?
1 Like
I am also looking for the answer to this question? Did you find an answer?