Hey. Great site. That's my converter you're referencing. Thanks for that.

Re accuracy, there are a few problems. First, xiao'erjing was never really standardised. You'll find multiple variations between different texts. Second, there is not a proper way to write -ng in arabic. In the case of Uyghur or Persian, the letter ڭ is used. When transliterating foreign words into Arabic, a combination of ن and غ, i.e. نغ is used. Xiao'erjing however uses the final mark ٍ (two small dashes below the baseline, pronounced "in") for both -in and -ing. thus 零 and 林 would both be written لٍ. In/ing endings are not the only case of this ambiguity imposed by the Arabic script.

Much of the data used for the conversion came from the wikipedia article and then checked against whatever examples I could find elsewhere. I've also taken some liberties when multiple options were available in order to clear up some ambiguity in the results but not so far as including -ng which was never part of it in any version i've ever seen.

Basically, it's accurate because it's a phonetic transcription that has never been standardised, and anyone who knew the Arabic script would be able to know what was being said as well as someone knowing the latin alphabet would be able to make sense of pinyin.

I've got a newer version of the conversion script that has greater support for tones and variant pinyin. It should be up there in about a week or two for anyone interested.