[GCompris-devel] Imageid

[Bruno Coudoin]

Le 17/01/2015 21:55, Karl Ove Hufthammer a écrit :
> On 17. jan. 2015 19:59, Holger Kaelberer wrote:
>> Maybe we should package the translated word-lists differently (then). 
>> Now they come in the voices-XX.rcc, which makes sense when word-lists 
>> come together with the corresponding ogg-files. Already for german 
>> this will not be the case as we don't have ogg-files yet (maybe same 
>> for word-lists that will be translated the po way). So it might make 
>> sense to package the contents.json in gcompris-qt directly? 
>
> Yes, I think so. Actually, I find the imageid activity more 
> interesting and challenging without the voice files. With the voice 
> file, you only need to read the words (and don’t really need the image 
> at all).
Hi,

Well it depends on the children reading experience and what you want to 
train. It can also be used to learn a foreing language and in this case 
the voice is important.

If you look at the 'lang' activity in the Gtk+ version, it is based on 
the same data set but includes several exercises at the end starting 
from the easiest with the voice to the more complex where you have to 
write the word. Also 'lang' included a first step where you see all the 
images in a set before going to the exercice.

I don't know yet how to organize all we can do here. Doing several 
activities or allowing configuration of the exercise. Do we keep 'lang' 
and 'imageid' or merge the two, I don't know. Input is welcome.

One important feature also is to let a teacher select the words he want 
to teach from our list.

>
> Without it, you have to look at the image and figure out which (of 
> five) words fit the image best. Of course, this is easy for very 
> concrete words and objects, like «apple», but for more abstract 
> concepts, like «scared» or «happy» or «friends», it’s more of a 
> challenge, especially for kids.
This is why the step with the intro of the set of word to learn maybe a 
good approach.
>
> BTW, that reminds me of another nice activity in the GTK+ version, 
> where you have to pair up related images (e.g., rain and an umbrella, 
> or a pencil and a postcard). In later levels it gets more difficult, 
> and in at least one level there are two reasonable matches for one 
> image, and you have to choose the correct one based on which of the 
> two also fit a third image. If this activity was ported, we could 
> probably reuse many of the art4apps images to create new levels. Some 
> ideas: process based (cow & milk, tomatoe & ketchup), antonyms (happy 
> & sad, rain & sun), related in time/season (dream & moon, winter & 
> coat, sun & bathing suit). It would be fun to try to come up with some 
> really difficult (but logical) matches for the later levels. :)
>
Yes, the images are a very important resource to create activities.

I have good news on that, Isabelle from art4apps contacted the original 
graphic artist and the images are done in Adobe Illustrator. As it is 
vectorial we can export them in SVG. It will take time as this must be 
done one by one as the images are mixed in a files by 10 images and not 
aligned.

Bruno.