The pronunciations are sometimes wrong, which is a bigger problem. Here’s an example:
Вчера, возвращаясь с работы, я нос к столкнулся со своей бывшей подругой.
The pronuncation field says that it’s носу́, but Wiktionary, and my Russian-speaking source, say that носу́ is the locative singular, while the dative (which is what we want here) is но́су.
Are prons added automatically, or manually?
From where are you typically clicking the pencil icon? If you open the game settings (button with a gear icon bottom right on the web), you should be able to turn on pronunciations which will show them automatically after answering a sentence. The pronunciations are also shown if present when managing a collection.
The pronunciations are machine generated. Please report any issues you might come across like this sentence and we can get them fixed.
I didn’t realize that. Now I’ve changed the setting. Thanks!
Does the pronunciation generator know the base (nominative) form of a noun or adjective? Is it able to understand cases, gender, and number? Very frequently, a word has the same surface form (spelling), but a different accent pattern, when it appears in different places in a declension paradigm. If the pronunciation generator can’t figure this out, it’s not good enough on its own, and there should be native speakers checking the final pronunciations. In fact, there should be native speakers doing this check anyway, since it’s clear that even if the pron generator has some understanding of grammar, it’s insufficient. Remember, too, that the wrong pron is worse than no pron, even though prons are extremely important.
Classes of errors in the pronunciation field of Russian sentences
I find it very helpful that the pronunciation field for Russian pronunciations is filled more often than not, usually with correct information, and that I can choose to always display this field after pressing the “Submit” button. However, errors are still quite common, and I thought it would be useful to classify the scenarios:
(1) There are two possible places to put the stress, depending on the case and/or number, and the automatic accent marker picks the wrong one.
(2) There are two possible places to put the stress, depending on the case and/or number, and the automatic accent marker chooses neither one, leaving the accent mark missing entirely.
(3) There is a “ё” character (always accented, and pronounced “yo”) that is written in the sentence as “е”, a common practice in written Russian that is not designed for language learners. In the pronunciation, it is written as “е” (pronounced “ye”, and capable of being either accented or unaccented). In this case, it’s not clear where the accent is supposed to fall, and whether that character is a “ё” or a “е”.
(4) There is a “ё” character (always accented, and pronounced “yo”) that is written in the sentence as “е”. It is written as “е́” in the pronunciation, which does show where the emphasis falls, but leads the reader to think, incorrectly, that the character is a “ye”.
(5) The pronunciation contains Latin characters, without accent marks. Beginners in their first stage of learning the language may think this is useful, because they’re struggling to master the 33 characters and are generally not even trying to put the stress on the correct syllable, but anyone who has gotten past that stage will find this useless.
(6) The pronunciation field is empty.
I think targeting these categories one at a time could be the best way to proceed. Ideally, you could produce a list of words with multiple possible accent patterns to help you identify sentences that belong to category (1) or (2). But I’m not sure how best to get it. I did find a Stack Overflow post about retrieving information from Wiktionary, though it seems that doing this is not simple. I’m not surprised, because in my searches, it’s hard to predict beforehand whether searching for a particular word will take me to a page for that particular (inflected) form, or to a page that tells me I need to do a full search for it, in which case I need to go to the declension table to find the accented forms. But however you proceed, you will need speakers with some knowledge of the language, and some of the work they do will need to be manual.
I am already reporting problems and updating my “Notes” field every time I see a problem. If there was a way for me to edit the “Pronunciation” field and have it go into the main database, I’d be glad to do it for you. I would only want to deal with the cloze word, though. Adding the accents for all words in the sentences would be too much work for me, so I would not want to tackle sentences in category (5) or (6) unless simply inserting the cloze word with Cyrillic+accent was acceptable.
This is great @alanf_us - thanks for putting it together!
Good idea, we’ll see what we can come up with, and thanks for the link. I’ll get back to you on helping moderate. Please continue to report any issues you come across in the meantime.
For 5 - might you be seeing these for pronunciations you’d edited, or perhaps for custom collections? Same for 6 - all non-custom collections for learning Russian from English (Fast Track, Most Common Words, etc) appear to have a pronunciation field present.
Will keep you posted. Thanks again!
Thanks for the response, @mike!
Please note that my knowledge is based on reference materials (especially Wiktionary), since I’m not a native speaker.
None of the sentences I’ve played so far has come from a custom collection.
I’ve never removed a pronunciation, and whenever I’ve tried to edit one, the changes don’t persist, so I haven’t done it more than two or three times. On those occasions, I always added the pronunciation in Cyrillic characters.