Thanks for the info. I tried again last night and the same MP3 transcription worked!
There is a very curious feature though, which I have seen also when experimenting with Whisper on my Mac: certain words get transcribed into Mandarin rather than Cantonese, typically very common ones, so 我哋 (o dei, spoken) becomes 我们 (women, transcribed), 喺 (hai, spoken) becomes 在 (zai, transcribed) and 同 (tong, spoken) becomes 和 (he, transcribed). It’s as though it picked out the meaning and translated it. From a Cantonese learner’s perspective, this is not what I want. Ideally the text should match the audio, It’s also the challenge with a lot of online materials, where Cantonese videos have Mandarin subtitles.
Another bug that has been recently introduced: when I edit the text in sentence mode, to correct these errors, I don’t see the updates after saving the edit. I have to exit sentence mode and come back. This used to work correctly.
Our current understanding of this issue is that it is due to the limited set of data to train Cantonese on (compared to Mandarin). The incorrect transcription is being done by Whisper as you noted, and it isn’t something we can easily fix on our end.
Right now the easiest options are to either continue to allow users to transcribe Cantonese with Whisper or to disable this feature for this language - If you feel strongly that the negatives of Whisper outweigh the benefits for Cantonese, let me know what you think. Otherwise it seems best to keep the feature enabled.
Regarding the edit sentence issue - I was unable to reproduce this. Could you provide more info? What device are you using and does the issue occur every time?
Clearly, it is better to have the whisper transcription available, so please keep it! Hopefully updates over time will improve the accuracy. In reality, correcting the text is, in itself, a useful activity, as it requires very careful listening. It just takes more time and more language experience.
As for the edit issue, yes, it happens every time. I am using a Samsung S24 Ultra, so the latest Android API version.
I noticed that there is a new statistic popup box but I miss the full information (words learned, LingQs, LingQs learned). Also I used to add my speaking hours manually and I don’t know how to do this now.
Hey @JulianiTOgo the statistics dropdown was simplified to focus it on activities that are directly within a learner’s control - Words Read, Hours of Listening & LingQs Created. The other stats can be seen when you click to “View all statistics.”
Also on the larger statistics (Profile) page, you can manually add speaking hours. Click the + button next to the Hours of Speaking stat. Although there is a bug on Safari browser which doesn’t allow clicks, and we are actively working on a solution to this.
New here so forgive me if I’m speaking out of turn … I am learning (simplified) Chinese and come across this issue as well. @mark Can we edit the transliteration? or can this be added? I find some of the audio doesn’t seem to match as well. though I am as new as one can be, so it could be my ear. @WillowMeDown do you find the audio is correct? (thanks for your corrections in the popular meanings:)
browser/ios/android : When we move between tracks in playlist, or between lessons, could the play head (audio and reader) be reset to the beginning? Currently it plays from last used location which is quite annoying, having to rewind - or only hear half the audio in playlist.
Just a thought … shouldn’t hitting previous track move the playhead back to the current track start, and then a second time should move to the previous track?
All audio files on LingQ remember their last location when you play them next. This happens on the lesson page and in the playlist. This was particularly useful when we tracked audio listening time by completed times listened. Now that we track all audio sessions independently we could look at that behaviour again. Having said that, there is nothing wrong with picking up where you left off in your listening. If you are listening to content like podcasts or audiobooks, you want this. And, having to figure out where you are in a lesson is a good exercise when listening.
Ah I appreciate your point, but after working through the sentences piecemeal in each lesson, it’s quite annoying having to go through the playlist once hearing just the “New words:” sections.
Any chance of a button to reset all tracks? Might this impact the accuracy of the statistics? Is that a problem? I’d be happy to give that up given the time I would save. Ahhh - do the statistics control the coins/points or whatever it is we get? I understand I might not get my quota
Could you explain what you mean by “hearing just the “New words:” sections.”? That sounds like it is specific to a certain course or lesson. That doesn’t sound like our typical lesson or course which usually has audio matching a transcript. We don’t usually have wordlists at the end.
Ah, it may be specific to the HSK1 (beginner 1) mandarin course. Each lesson consists of four sentences, followed by a new words section that repeats the new words used in the lesson.
So I added the course to a playlist (which is brilliant!). I go through the playlist until I hear something I don’t understand. I use the open lesson option (from the playlist, brilliant again) and check the sentences. Then I use the 3 dots drop down menu from the lesson to go to the next lesson. Go through the lesson sentences … and repeat. Also worth noting, if I return to a lesson, the progress bars at the top are at the end, where I left the lesson previously, which as you mentioned maybe desirable in some circumstances.
Later, I go for a walk and start the playlist, but the lessons I worked on piecemeal start at the new words section, not the beginning of the lesson.
Hope that clarifies the issue. Thanks for your time, Nick.
Edit: just experimenting, Some lessons have the new words section transcripted, not sure if that is a recent update. If I go through to the end of this, then go back to the playlist, it skips over that lesson completely.
Ok, that makes sense. It sounds like you really don’t have much use for the wordlist which is what is causing the problem for you. Why don’t you find another course? There are lots of courses to choose from. We always push the Mini Stories since they repeat the same stories from different perspectives and finish with questions at the end. Not that you need to answer the questions but it’s the repetition from different perspectives and in question and answer form that is so effective.
I started with the mini stories, really good by the way, excellent quality audio, but was struggling … then I noticed they were beginner 2. So, I figured HSK1 would be ideal … but again I take your point, I will have a look around, thanks.
Mini Stories do say beginner 2 but really should be fine. They will just take a bit more time at first. But, you will see that you gradually pick up steam there. Otherwise, there should be lots of beg 1 content that doesn’t have the issue you are trying to avoid. Good luck!
I found a new feature. I sentence mode there is a magnifying glass bottom right in the app. When you click it you get a series of exercises: flashcard matching, sentence reconstruction and a reading out load test. The reading test is not working for me. The sample audio is in Cantonese, the text to read in Chinese characters but when I speak it transcribes into random English and gives me 0%. For sure my pronunciation is not the best but it seems like it is maching in the wrong language.
