LinqQ audio differs from lesson audio track

Similar to issues I reported here:

The lingq audio (Login - LingQ) should be pronounced “おおぜい” in the context of this lesson

Same issue …

The lingq audio 表して いる(Login - LingQ) in lesson (NHK Easy 2017年・秋冬, 東京オリンピックの記念にナンバープレートを作る) should be あらわしている in the context of the lesson.

HI sjordan,
Are you saying that lesson audio is incorrect or TTS for highlighted word?


Given the fact that many kanji (like 大勢 and the 表 in 表す) can be read in different ways, I don’t see how you can expect the automated pronunciation to always be correct.

@amop567 I realize there are different pronunciations.

However, as a developer/designer myself, I try to alway remember the Principle of Least Astonishment (POLA)(Principle of least astonishment - Wikiwand). Essentially, how would the average user expect a system/application to behave.

Given that one of LingQ’s main guidelines is interact with the language through natural exposure of listening to the audio and reading along, I think it’s fair to assume that the vocabulary (TTS) would match the audio dialogue.

Steve has made the point in numerous places on LingQ and Youtube to first use the audio and reading as a primary language aid. All the other stuff, SRS, vocabulary lists, grammar etc. is secondary to LingQ’s style. I believe Steve has even advocated not bothering to learn all the pronunciations of a word when you first see it; rather learn the pronunciation as it’s used in the lesson. Then in the future, in different lessons, the other pronunciations will be reveal themselves. I think this is the point of why in keeping with LingQ’s owns philosophy, the vocabulary (TTS) should match the audio transcript.

Now, having said that, i did not realize before zoran asked, that the vocabulary was fully TTS (meaning that LingQ does QA the results against the actual lesson).

If LingQ doesn’t have the resources to make that a reality, we can acknowledge that. But again, philosophically and in keeping with POLA, I don’t this it’s an unreasonable assumption that the vocab and the audio as read in the lesson would match.

Not a developer so I can’t say for sure, but I don’t know how they could guarantee 100% accuracy without manually checking (and correcting if necessary) the automated audio of each word in all the articles on the site. There’s too much ambiguity in the kanji readings for an automated system to work absolutely perfectly.
Of course it would be ideal for the automated pronunciation to always be correct but I just don’t think Lingq has the resources to do that (assuming it’s even possible).

I think it comes down to being impractical and not enough resources to check the vocab audio. It’s up to users to point out inconsistencies.

Which makes this thread valuable. It’s the feedback loop built into how LingQ operates. It’s essentially community outsources QA.

So, I’m reporting an inconsistency in this instance and now it’s over to LingQ to update the vocabulary in question.

Presumable when someone comes across another inconsistency, they’ll do the same and incrementally the quality will improve.

1 Like

Yes, it helps, but only to an extent. Maybe next time someone will come across a reading where 大勢should be pronounced たいせい. These issues can only be solved on a lesson to lesson basis, so I don’t think we can expect any big improvement in the overall quality of the site in this regard. As you know, there are not enough set rules for automated pronunciation to work perfectly. Even google maps will mistakenly pronounce Japanese place names from time to time.

I agree. It has to be lesson by lesson.

It’s great that you mention Google maps because Google does exactly that kind of community based QA with Google Guides.

I’ve contributed to Google Guides. If you look up a shop or restaurant or something on Google, you may notice a link for “edit” (or something similar). If you click that, you can update the listing with things like “shop is no-longer in business” or “business hours are wrong”. I have done that when I find inconsistencies in Google Maps.

Same principle here with LingQ lessons. We users find inconsistencies; then if LingQ is committed to quality, they update the lesson appropriately.

PS. Thanks for the rose