To just learn characters without using them is not a good practice so I try to read as much as possible. Many years ago i did write a program that parsed Chinese texts i had in a database and gave me a list with texts and the percentage of known words in that text. Then I could choose suitable texts for the level I where at for the moment. In the texts words/ characters that where not known by me where linked to a dictionary and also had a translation popup etc. It would be great to be able to use something similar in skritter. Maybe just list example sentences with many characters/words i do know. This is most suitable for beginners/ elementary learners with a limited vocabulary but at that stage i feel that it is very important to really use the language an read a lot. For now I will try to find my old code and dust it of;-)

Seems like an interesting idea, if only to basically call out characters you don't know.

I've been experimenting with code to pull example sentences with a large percentage of characters you know, so as to learn the ones you don't and have them in context.

Precisely, that is what I have been doing as well. I where harvesting Chinese newspapers etc for suitable sentences and texts. I have a screenshot of what i did on my own: http://i973.photobucket.com/albums/ae220/mandarinboy/textreader.jpg Machinese translations is not always 100% perfect but they do give you at least a good start. It is the actual reading I am interested in. To read i need to have texts that i do not have to look up every single word in ;-) Over the years I have forgotten way to much Chinese.

I've had that feature in mind for a long time, but we still haven't implemented our own text segmentation algorithms, which we'll need for it. I personally would love it, but I don't think we can get to it any time soon.

I wrote a small program to do something similar but not for texts, just sentences. I use the example sentences that come with Mnemosyne (20000 sentences). The sentences are in simplified. I had to convert them to traditional first.

The sentences are here:
http://www.mnemosyne-proj.org/node/115

What I do is that I export all words/chars I know from Skritter. I then select sentences that contain only characters I already know. I found that will also yield words I do not yet know but that consist of characters I have learned though. I then try practicing these sentences in Anki. I do not practice the pinyin as presumably I will already be able to read all characters.

Interesting. By the way, those are also the same sentences we're starting with as examples.

For text segmentation have you considered Adsotrans libraries? Supposed to be open source, maybe the license will work for this.

It is open source: http://adsotrans.com/downloads/
The developer for that actually did go over to Chinesepod to work for them. I think that they are using that tool as well.

We need something pure Python so that we can run it on App Engine. I think Adsotrans is in C.