Greetings, Comments inline...
> From: Tony37 <[email protected]> > Date: Mon, Apr 20, 2009 at 9:55 AM > Subject: ImageQuiz > To: FGrose <[email protected]> > > > Hi, > > ImageQuiz uses a cvs file format for questions. I have added an XML file > format. Currently, both files are used to load questions > into an sqllite database. Is this CSV or XML format documented anywhere? (http://wiki.laptop.org/go/ImageQuizPlus doesn't seem to have much) Secondly, how attached are you to the file format; that is, would a more general file format be acceptable (supposing it supported the necessary components). > For me, the problem is figuring out how to interface with the > journal/datastore. In a hierarchical file system, quizzes would be sets of > questions > in XML or CVS format within categories determined by the folder names. An > imagequiz_library would be a base folder containing these quizzes. This > organization does not map well to the journal/datastore at present. In time, > there will be a journal objectchooser which can show entries of the proper > mimetype in the journal or in specific repositories on the schoolserver, usb > keys, etc. The design we (the RIT student team working on a question API) have been working from eschews the Journal entirely, focusing on a URI approach to locating lists of questions. This could conceivably include specialized URI schemes like "journal://uuid", but can also include very straightforward access mechanisms like "file://" and "http://";. Using a URI-based scheme allows query parameters to filter/select specific types and categories of questions; I'm still undecided as to whether something more sophisticated for question filtering/selection is needed. > Another problem is whether the unit should be a set of questions or a > question. This is similar to the issue of where to store an image or mp3 > file and how to organize a set of them. Perhaps the journal could have an > album where a single question (or image or mp3) could be located by tag > (album name) by looking in many different albums (similar to picasso for > example). We've, so far, been treating the question as the fundamental object, essentially ignoring the grouping problem by making it external to the API. This is easier to do when a specific URI is required to load question data — but also inhibits automatic discovery of question content (without explicitly providing an additional mechanism to do so) as there is no central way to ask about "every question available". You could query a list of potential URIs for questions, but that's probably not the best way to go about it unless you know the URIs have questions or the list is small. > The implementation I am working on will find sets of questions (called > quizzes) and map them into categories based on the folder structure. The > base will be 'imagequiz_library' and, hopefully, there would be one in the > activity bundle (built-in samples), on the schoolserver, and on removable > drives. From the API design we've been contemplating, the usage would follow something like this: - Sequentially open a series of URIs for the various locations questions might be located, suppressing errors for missing or unavailable sources. The URIs would either use a format parameter that specifically parses and builds question objects according to the hierarchical scheme you outlined above, or some other format supported by the API. for example: q =quizdata.open("file:///path/to/local/questions/?format=image_quiz") q.append( quizdata.open("xmpp://quizmas...@school-server/quizname?format=generic_xml")) - In the activity do additional filtering and combining of the resultant lists of questions to form the full question set that comprises all the sources. - At that point you could push the discovered questions into a local database (if that makes sense) along with the additional processing (retrieving image URLs, etc) needed to present the questions offline. Or you could just present the questions, retrieving additional resources as required in an online fashion. > The focus of ImageQuiz is matching an image or a place in an image to the > answer to a question. For language learning, I am setting up a question bank > where the question is an audio clip (e.g. 'ball' spoken in the target > language), and a set of pictures, one of which shows a ball (or a picture of > several objects where the answer is to click on the ball). ImageQuiz will > also support standard flash cards (text question, text answer) plus > combinations like audio question, text answer. Sadly, speech recognition > isn't ready to support audio question, audio answer. At the most basic level, this should be straightforward enough to accomplish; however, managing the structure and format of multimedia (image, audio, video, complex/multilingual text, etc.) questions requires some careful thought to be flexible, simple, and general enough to be widely useful. The details will need to be fleshed out; hopefully we can find a common base so neither of us reinvent too much... Greg S. -- Email is packaged by intellectual weight, not volume. Some settling of contents may have occurred during transmission. _______________________________________________ FourthGradeMath mailing list [email protected] http://lists.sugarlabs.org/listinfo/fourthgrademath
