I'd love to see Duolingo bridge the gap between the knowledge-tree-based lessons and translations. One way to do this would be to select text to be translated that contains only those words and grammatical rules that have been introduced within the knowledge tree.
Another way would be to show a score next to each translation--identify the amount of overlap between words/phrases in the current user's vocabulary and the words/phrases in the text to be translated. It could be as simple as providing the number of words "in" and "not in" the user's vocabulary.
"This text has 814 words in your vocabulary, and 372 words that you haven't learned through our lessons." ...or simply "(814 of 1,186 words known)"
I imagine that users here learn many new words while translating text due to the need to look up words not yet learned. Duolingo could scan the user-translated text, and add those words to the user's known vocabulary. For these new words, greater strength scores can be given based on how often they appear in the user's other translated texts.
"Duolingo could scan the user-translated text, and add those words to the user's known vocabulary."
I have read elswhere that DL adds this words on you vocabulary.
Your suggestion to bridge the gap between the knowledge-tree-based lessons and translations is very welcome.
User sommerlied mentioned the site bliubliu.com in another post (http://www.duolingo.com/comment/912265). This site does a great job of tracking which words you're familiar with. It gives you some text in your target language and lets you underline words you don't know by clicking or tapping on them. Going through a series of of these, it quickly builds a list of your vocabulary in that language and then presents you with content that closer to your level of understanding. It also offers drills to help you improve.
Perhaps the translations in Duolingo could incorporate some of these ideas, making translation a more interactive and didactic exercise. If we could offer hints about which words we didn't know at the outset of a translation, it might offer clues to others who need to check that text and verify it later. How confident was the translator? Did they already know most of the words or did they look up nearly everything? And this could feed back into lessons somehow. Just a thought.