I appreciate that translation can sometimes be as much of an art as a science, and one of the nice things about Clozemaster in general is that it has sentences written by a lot of different people who will each translate things as they see fit without any pressure to adhere to certain standards or guidelines. This allows you to see different ways that words can be used and different ways to say a single thing. This is a good way to learn a language.
However, the purpose of the Fluency Fast Track is to learn each word from only a single sentence. From my point of view, if I’m going to spend a lot of time repeatedly doing a single sentence in order to learn a word I’ve never seen before, then it’s important that this sentence be a good representation of how the word is most often used. I don’t know which sentences you say are colloquial translations, but as far as I’m concerned the sentences I reported either had unequivocal mistakes in them or were not very accurate translations – some to greater degrees than others.
My teacher is from Lviv. But I don’t think that should matter too much because I only asked her about the translations if I was still unsure after checking both the dictionary and sample sentences on globse.com
If I had confidence that the sentences were professionally translated then I’d probably be a little less trigger-happy on reporting. But I mean. There are literally Russian words in there. And a simple word like шиє was misspelled in a way that would make sense to a Russian speaker. Certainly Ukrainian is spoken differently amongst different communities and in different parts of the country – and one part of that is because Ukraine is a bilingual country and many people don’t speak Ukrainian correctly. I don’t want to learn Ukrainian the way it’s spoken by a Russian speaker – I want to learn each language separately and correctly.
The obvious mistakes themselves aren’t the problem. The problem is that when there are obvious mistakes, then there will also be subtle mistakes that I don’t notice. I reported a lot of obvious mistakes when I did the Random Collection, and I’m only just now learning that many of the other words and phrases I learned in there are actually incorrect as well (mostly all from a single Tatoeba contributor). I don’t have any regrets about that, because I learned a lot of stuff really fast, but I’m at a different point in my learning curve now and I have the luxury of being able to just choose not to spend time with material that doesn’t sit right with me.
It’s so sad… I’m really looking forward for the Arabic Course, the 6th most spoken, but no signal yet…
But is very good to see so many languages ready to launch or in progress!
@Roberto.Melgaco: I think you’re referring to the searchability of the “new” FFT. I have complained many times ever since it was introduced, but sadly it hasn’t been implemented yet.
I second @morbrorper 's opinion.
Unsearchable collections through the universal search bar are not only the new FFT but also personal collections. I’ve (indirectly) complained about this issue but no response from the admin yet.
I guess the overall data architecture was poorly designed and Clozemaster cannot quickly fix the issue, given the previous comment from Mike (the admin) on the data management. Once I proposed a tagging system and he said:
…how to actually migrate everything if we wanted to change the underlying architecture, etc. It would indeed be a long-term reform like you mentioned MsFixer, but there are some worthwhile potential benefits like you mentioned as well.
His comment implies that each collection is separately managed and there is no single repository. Therefore, massive data migration is required if you want to use the universal search for a sentence in the new FFT or your personal collection.
We’re aiming to start on Arabic next month, will update accordingly.
We have a backlog of French, Swedish, and Turkish ready to go at the moment. Italian has completed another round of proofreading, updates will be made soon. Ukrainian is going through another round of proofreading. Russian, Polish, Lithuanian, and Croatian are in progress.
@Roberto.Melgaco@morbrorper thanks for the nudge on this request! It keeps getting bumped by other projects/features/fires, but it’s on our list. I don’t have a specific timeline yet. All that said, it’d be helpful to understand - why are you using search? What are you trying to do?
I assure you the overall data architecture was expertly designed. Even “quick fixes” take time to do right, and other projects simply have higher priority at the moment.
I come across a word and I want to learn it, so I search for it in the corpus.
Somebody comments on a sentence and I want to see the whole picture or hear the audio, maybe file an error report.
Looking for synonyms: I often use “Translation” search to see what Spanish words are used for a word in English.
Searching for patters: what preposition is used with this verb?
Of course, only having the “*” wilcard is a severe limitation. And, an option to limit the search to only return the sentences I’m playing would be very helpful, as it would save me the work of creating a new cloze when I already have it in a collection somewhere.
These are all very good reasons listed by @morbrorper.
If you want to go to a certain sentence but don’t remember or know in which collection it is, you currently have to try out all collections individually (100 most common words, 500 most common words, 1000, 2000, and so on).
I don’t think it’s that bad though. The “legacy” collections are globally searchable, but not the new FFT, nor the private collections. At least in Spanish from English.
@mike
Sorry if I didn’t sugarcoat my previous comment enough that might get you emotional. But I have never seen the following issues in other servicers/apps. Those issues are largely attributed to a failed architecture of “hierarchical database model” and/or “hard-coding” without considering future expansion.
Cannot search sentences across collections.
Cannot use even simple regex to narrow down search results.
Cannot edit and save sentences, translations, notes and hints.
Cannot “relocate” a sentence from one collection to another at once (i.e. you need to “copy” it to the new collection, and then “delete” it in the old one.)
Cannot tag sentences.
Maybe some people might tolerate those issues if it remained as a free function. But sentence search, for example, is a premium function for PRO users only. Even Anki offers these those functions with no errors for free.
Before adding new “seems-to-be-fancy” but “half-baked” features, I do hope Clozemaster to re-review previous user error reports and requests for improvement.
The wildcard only works at the start and at the end of a word, not in the middle. Ideally, I would like to be able to use regular expressions to find patterns, like verbs and their associated prepositions.
There actually was a time when we could use regular expressions, but that was unofficial and undocumented and was made impossible with the introduction of some other feature.
Unfortunately, I do this a lot, cause I am always typing it right - I can hear well and type fast - but thinking again and resetting the sentence for not be so sure what it means anyway.
I appreciate your active engagement in the forums, Mike! The transparency is especially wonderful. I would love to know if you can tell us anything on the specific timeline of the French fast track rollout. I have plans to learn German through French in a few months, and what I’ve heard from the new fast track this seems like an excellent medium for it. Thanks again!
I gave a glowing evaluation of the new Fast Track earlier in the thread (on April 15):
At that point, I was thrilled that the new Fast Track was designed to let you (1) jump in at one of ten levels of increasing difficulty and (2) only have to encounter a word form once (per track, I think). I hadn’t played long enough to determine whether the collections actually lived up to the promise: was each level really more difficult (composed of significantly less common words) than the previous one, and was each word form actually presented only once?
As time went on, I became conscious of some repetition. However, I haven’t been able to keep track of whether I’m encountering multiple forms of the same word or the same word form, nor whether I see the repetition within a single track or over multiple tracks. Trying to remember this would be an unacceptable cognitive load on top of the main activity of filling in the clozes. Normally, if I encountered a suspiciously familiar cloze word, I would rely on search to determine whether I had actually seen it (or another form of the same word) before. But the fact that search doesn’t work yet in the new FFT means that I can’t do that evaluation.
So one of my answers to “Why do you use search?” is “To determine whether or not I’m crazy when I think I’ve already encountered the same word.” Another is “Even if the word form has appeared only once, how many marginally different forms of the same word does the collection contain?”
To be clear: this is especially important for a language that I find easier, since I want an efficient way to go through relatively large numbers of words that I know in order to find the occasional ones that I don’t remember. For languages that I find more difficult, repetition is not as annoying.
