a few problems I've noticed
Some of the questions are so simple that my memory isn't being used to answer. In questions where you pick from a set of words and put them in the correct order:
- Sometimes there are so few words that there's no challenge.
- Other times, the capitalization of the first letter on some of the words provides a strong hint. I find myself eliminating words simply because of capitalization, and the remaining set of choices is then too easy.
- Occasionally, the only reasonable choices are between two words such as "a" and "an", making the question more a test of my native English ability rather than the language I've chosen to learn. This is the case in multiple choice questions even more frequently.
In multiple choice questions:
- Three choices are not enough to be challenging.
- I am often provided three possible answers that, although they represent my current level of grammatical proficiency, differ only by their use of simple words like apple, cup, or woman rather than testing me on more appropriately complicated matters of grammar that I need help with.
The last problem I'll mention here is that I'm questioned on words I haven't encountered yet in ways I'm not prepared for. This isn't a big deal to fail and them come back around to get right the next time, but it would be nice for Duolingo to have the intelligence to introduce words via questions in a better order. Some of the questions do this well, allowing me to tap on a new highlighted word to see it's meaning.
I completely agree with everything you said. There was a discussion on the same topic earlier this week that you might find interesting: http://www.duolingo.com/comment/882216
One of the most interesting ideas that came up in that discussion was the idea to use (carefully selected) incorrect answers frequently submitted by other users.
Mosey on over from you phone or tablet to the desktop version and it does become a bit more challenging as you will never see a set of words. I find the tablet / phone version quite a bit easier. I think they have the word selection so people don't have to type it in. On a tablet it's no big deal, but a fat-fingerer like myself appreciates the set of words on a cell.
I don't think anyone's saying that there can't be problems where you select the words from a word bank. I understand the need to limit the amount of typing required of someone on a phone or tablet. But the way the word bank currently is, you usually only get 1-3 extra words and the first word of the sentence is capitalized. 9 times out of 10 you could come up with the correct answer without ever looking at the question just by looking at the word bank.
These problems could be made a lot harder simply by adding in more words to chose from and removing the capitalization from the first word of the sentence. No, it still won't be as hard as creating the sentence from nothing, but it will provide a much greater challenge. I cover the word bank until I derive the translation from scratch to make it more challenging, but I hope they will use some of the ideas from this thread when they update the apps.
I also imagine that lot of people using the phone/tablet are using it because that is what they have available during the time they have to practice. Depending on their schedule, this could mean they primarily or even exclusively use the app and may not have the option to just use the website.
Some of the cited problems can be partly solved by letting us choose between the different types of questions:
bvb only multiple choice, or only translate to the language to learn, or only translate to native language, or only choose between images, or a combination between the different exercises.
You would then practice first with choice of pictures for the new words in the lesson (and you don't have to speek for them) , then multiple choice, then choose from given words and in the right order, then translate it to the native language and then translate it into the language to learn and finally write or translate spoken sentences. This will build your level of difficulty and let the student free choice.
For the exercises with choice of different translations I have a note.
The possible choice of grammatical wrong phrases should be avoided, because you read this and the sentences are included in your memory. It therefore seems to me choice of incorrect answers from other users is not appropriate.
Learning the correct preposition, or genitive of the definite article, ... can best happen with one sentence, with a top down menu for the choice of the possible prepositions or articles. Choice of sentences containing incorrect prepositions or articles seems to me not pedagogical.
All the phrases shown must grammatically correct.
While I agree it would be nice to limit the types of questions delivered in a learning session (thereby focusing on a particular skill or skill level), ultimately I think each question type has issues and quirks that need to be corrected to improve the quality of the training system as a whole.
Here's a specific example from Italian.
The question: "Quali animali mangiano lo zucchero?"
Words provided to construct the answer...
From the correct answer: Which animals eat sugar
Other words offered: beer cheese pasta Fish alcohol wine oil Coffee chicken egg
Assuming at this point that I've mastered the vocabulary words and having each of their strength indicators up all the way, what I really should be learning here is how to ask questions. What kind of question would I begin with the words Fish or Coffee, since those are the only other capitalized words? A better selection of words for me at this point would be other interrogatives: quale, questi, questo, quando, chi, che, etc.
This is why I think it's important to track grammar rule memory in addition to (and separately from) vocabulary memory. I'm being drilled way too much on vocabulary I already know, and not enough on ways of putting sentences together where I'm actually in need of more training.
That's an interesting concept, tracking grammar rule memory. I don't think Duo is really designed to track thousands of grammar rules. It's designed so that the learner has to discern grammar rules from the exercises much like Rosetta Stone. That can get frustrating for sure. It's really vocabulary focused which is why you may need to turn to other resources to really hone in on Italian grammar rules.
I use Butt and Benjamin's Grammar of Modern Spanish book and workbook in my Spanish studies (yes, I know that won't be of use to you) but something similar is probably available for Italian. Check out the Italian grammar book reviews on Amazon. Language learning books are generally inexpensive and will keep you occupied and learning for a long time! I find that having a grammar reference book with a companion workbook to be the best combo. Good luck on your Italian studies!
Duolingo doesn't need to track thousands of grammar rules. About a hundred or so should suffice.
I'm actually completely in agreement with the idea of inferring grammatical rules based on usage in the lessons and strengthening drills already in use by Duolingo. But notice how intelligent Duolingo already is when it comes to parsing your answer and detecting errors. If it fully used what it's already figuring out about which pieces of an answer you got wrong, it could update its knowledge about your understanding of grammar at the same time that it updates your familiarity of the used vocabulary.
Did you pick a first person plural form but missed that it should have been possessive? Did you use a third person pronoun instead of second person? Do you habitually confuse verb conjugation forms for -iamo vs -ano in Italian (2nd person plural vs 3rd person plural) because they look and sound similar? Are you selecting the wrong verb tenses for irregular verbs? Then how about recognizing these things and giving us exercises that will train us to correct them instead of treating everything as a simple problem with vocabulary?
Just as the most common words represent the great majority of words used in text or speech, the most commonly-confused dozen or two grammatical rules will account for the great majority of errors. If Duolingo can identify and track these--even if it does so behind the scenes without the users being aware that it happens (although I think there's value to making that visible to users)--then the system as a whole will be far better at getting us to practice the things that we actually need to learn.
Sure, eventually with an additional few hundred hours we'll absorb that grammatical knowledge anyway, and it will likely be through lots of less productive Duolingo training (because the spaced repetition algorithm is not targeting grammatical distinctions), and it will continue to require a focus on external means of language learning such as immersion with native speakers (a good route to go when feasible). The reason I bring this up is because the Duolingo product and team are already so progressive in the area of language learning, and it's been improving continuously so far.
Language isn't a flat list of unrelated terms, but an interconnected system of rules and patterns. Treating it as such in a training system will only improve it. How, or even whether, it's visualized to users is another matter entirely.