16
General Software Discussion / Re: Nova the Easy-Peasy Do-It-Yourself Text Aligner
« on: December 07, 2018, 08:24 AM »
@cranoscopial: I had not even seen that typo, but your discovery is just another hint at developers to run some spelling checker at the end (most Germans do so, and then their texts are full of "sie" instead of "Sie", and vice versa, "sie" being "they" and "Sie" being "your"...).
As for the work this aligner obviously isn't willing to do, of course there are dots, "full stops", at the end of sentences, but also after abbreviations, but after most abbreviations, the next word begins with a lower case letter; most sentences begin with an uppercase one, and as for ^r^n, ^r and ^n, which, especially for text downloaded from the web, create havoc for any script, I first normalize them to ^n, then run the script, then normalize them to ^r^n; any aligner should do so, too, instead of trying to cope with all 3 variants at the same time which is unsuccessful anyway.
Then, there are lengths, German and French being some 10 or 15 p.c. longer than the corresponding English text in most cases for example, so an aligner, while taking such differences into account, should compare character counts, too, before deciding if a dot is a full stop (new sentence; new visual "paragraph" on screen then if the users opts for that) or not; other simple tricks could apply, too, and then you minimize errors of such a tool, in this really simple task.
Btw, it's debatable if the systematic left block - right block paradigm is really optimized for every use case, I'd prefer the sentences being one beneath the other, then a blank line, the two "blocks" being distinguished by different color, but it's perfectly conceivable there be aligners who do it that way already.
But anyway, the developer in question obviously makes a lot of fuss around a (what's more, very simple) core task of any aligner, AND tries to "sell" that miss as an advantage of his tool, on top of this. Also, there is no mass market for aligners, so he faces prospects who should be able to see this them misleading of his, and that makes it all the less comprehensible even. ;-)
As for the work this aligner obviously isn't willing to do, of course there are dots, "full stops", at the end of sentences, but also after abbreviations, but after most abbreviations, the next word begins with a lower case letter; most sentences begin with an uppercase one, and as for ^r^n, ^r and ^n, which, especially for text downloaded from the web, create havoc for any script, I first normalize them to ^n, then run the script, then normalize them to ^r^n; any aligner should do so, too, instead of trying to cope with all 3 variants at the same time which is unsuccessful anyway.
Then, there are lengths, German and French being some 10 or 15 p.c. longer than the corresponding English text in most cases for example, so an aligner, while taking such differences into account, should compare character counts, too, before deciding if a dot is a full stop (new sentence; new visual "paragraph" on screen then if the users opts for that) or not; other simple tricks could apply, too, and then you minimize errors of such a tool, in this really simple task.
Btw, it's debatable if the systematic left block - right block paradigm is really optimized for every use case, I'd prefer the sentences being one beneath the other, then a blank line, the two "blocks" being distinguished by different color, but it's perfectly conceivable there be aligners who do it that way already.
But anyway, the developer in question obviously makes a lot of fuss around a (what's more, very simple) core task of any aligner, AND tries to "sell" that miss as an advantage of his tool, on top of this. Also, there is no mass market for aligners, so he faces prospects who should be able to see this them misleading of his, and that makes it all the less comprehensible even. ;-)