The mystery of 和

Hello,
I have a weird (small) problem : I cannot get the character 和 to be recognized, not matter how I write it, in which strokes order, big or small, thin or fat, impossible.
It could be that I am clumsy in the way I write it on the screen, but since all other characters are recognized - admittedly, somtimes after a couple of tries - I can't understand whay this particular (and not so complicated) character is never, ever, recognised by my Samsung Galaxy S2 (Android 4.1.2).
Any idea ?

That's odd - are you using the free handwriting recognizer or the Enhanced one? And what direction are you drawing the first stroke?

Hi, Mike,
Sorry for the delay in responding.
I don't know how to determine my handwriting mode.
However, I just tried to draw the first line from right to left, and...voila, now the character shows...I should have tried this first. I am very surprised because for other characters, I know that I sometimes make mistakes in the order or in the direction of the strokes, and I can always find the one I want among the options shown at the bottom of the screen.
For example bu (not) or bei (glass, cup) are recognised whatever the direction....any explanation ?
Sorry for wasting your time, and thank you for your support. It feels good to overcome this nagging obstacle...

Interesting... no matter how I change the settings, it always shows up as the first choice for me. Maybe you're doing the "box" (口) differently? The standard way is like that:

Yes, I write it like that ; all characters containing "Kou", that I write, are recongnised by the system...
And I have never faced this problem - apparently the direction of the first stroke is critical - for any other character (else than 和).

So when it doesn't recognize 和 the way you write it, what are the matches that you get instead?

I get for instance 你, 机, 相,租 ... there is some ressemblance, but these characters are quite different from 和 - my 口 is neatly closed, and for this reason, it should exclude 你, or 机 for example.
Also, I always do this 口 with three strokes only, as shown in your first post, so it should also exclude 相 and 租. For me, in a handwrirting recognition system, the direction of one stroke should be a much less important than the number and position of strokes, and less important than the final shape of the character.

I was just about to reinstall Pleco anyway, so before removing it I deactivated all my purchases and went back to the basic handwriting recognizer just to check again. Still, I really can't seem to be able to write 和 in such a way that it doesn't get recognized: even if I never lift my finger and draw everything in one go, and even if it's totally illegible (see screenshot), I always get it as the first or, at the very least, second match. So I guess I can't really help you with your question directly.

But if you want some unsolicited advice from a fellow Chinese learner, it's really a good idea to follow the proscribed stroke order. There are several advantages: it'll help you memorize the characters faster, you'll be able to see through the complex ones more quickly and notice the patterns even if you didn't know them beforehand, and it'll be easier to learn the new ones as well. Being familiar with the stroke order is the make or break factor when trying to read handwriting. Most importantly, it's almost always the fastest way to write.

Sometimes there's some discretion as to the stroke order, but that's not the case here. Since you're learning it (I'm assuming), why not just learn it the proper way right from the start? Even if you can get 和 starting from the left, you'll run into problems with some other character (I'm guessing 夭: you'll be getting 天).

ziggy said:

I get for instance 你, 机, 相,租

Click to expand...

You should be getting 10 suggestions if I'm right, have you tried scrolling past the first few? Maybe 和 is still there somewhere farther? Also, are the suggestions consistently the same, or just seem random and different every time?

ziggy said:

For me, in a handwrirting recognition system, the direction of one stroke should be a much less important than the number and position of strokes, and less important than the final shape of the character.

Click to expand...

Actually, it's totally the opposite, at least for Chinese. It doesn't matter what the final result looks like, as long as you keep the stroke order right. The opening stroke is particularly important, it pretty much decides everything, the last ones can usually be skipped altogether in the case of more complex characters. Here, I start seeing 和 as a result without even doing the last 3 strokes.

It might help to understand how the handwriting recognition works. Basically, every character can be written as a series of strokes it is composed of. For example, 木 is (h, s, p, n) [for 橫、豎、撇、捺] and this actually distinguishes it uniquely from all the others: even 朩 would be (h, s, p, d) [for 點]. When you're writing, what the recognizer does is (mainly) to convert your hand movements to a series of standard strokes and then look for the closest match in its character database, with some tolerance for errors. So the most important thing is to only do the exact number of strokes necessary, definitely not more. From my experience, how the strokes are positioned spatially is of secondary importance.

Plus, you also need to consider two other things:
- Handwriting recognition is for people who know how to write characters, and these people generally write them following the standard stroke order, which allows them to write fast. If there was too much ambiguity allowed, it would basically mean disregarding the writer's stroke order, which conveys a lot of important information. Then, handwriting recognition would become just like OCR and there would be too many mismatches when the stroke order was adhered to (as this information would be excluded from consideration). In the end, this would annoy the principal group of its users.
- Strict recognition algorithm is actually an advantage if you use the handwriting module for flashcard practice. Otherwise, even when you don't really remember the character, you can "cheat" and draw something vaguely similar (your best guess), and the actual character would then probably appear as one of the matches.

I'd say the handwriting recognition in Pleco is actually pretty good (from my point of view, at least). Sorry I couldn't exactly pinpoint when the issue is but if I were you and couldn't get a character recognized, I'd actually take it as a hint that I need to reconsider the way I'm writing it (unless the character is very obscure, then maybe it just isn't in the database).

Thank you for your detailed response and for your suggestions.

"an advice from a fellow Chinese learner, it's really a good idea to follow the proscribed stroke order."

Thank you. I actually try, of course, to learn stroke order, number, and position alltogether. I beleive however that number (and nature: i.e. dots, horizontal line, vertical line..) of strokes are more important than the direction of drawing.

I don't think I agree with you about the fact that a scribbling done with the right direction of strokes, but that doesn't look like the character is "better" for a Chinese than a character looking exactly as it should but written in random direction. The important thing once the character is written is obviously the result: what you read is actually what you see on the paper, isn't it ? Even written in dotted lines or even if you draw the strokes botom up, the shape of 人 is identified, whereas if it looks like 八, even if the strokes were drawn in the right direction you would mistake it for a 八...

Your screen capture of 和 identified from a very strange shape is impressive... as far as the technology of the software/smartphone is concerned. But I doubt that a human chinese reader would identify it at first glance as a 和 !

Only a software, or an observer watching from above your shoulder while you write can know if you draw the first line from right to left (unless you are using a paintbrush...). As long as a stroke is of the right length and position within the final shape of the character, only its length and position/orientation matter, not the direction of drawing. Printed characters have no "direction" of drawing at all and still are perfectly well understood.

To answer your questions :
- "have you tried scrolling past the first few?" Yes of course I did, and I was unable to find 和 among the 10 (or so) options.
- "Also, are the suggestions consistently the same, or just seem random and different every time?" Not random, the system suggest characters that are close to 和, but there are variations, based on how well or badly I draw the character on the screen.

To contribute to the analysis of the issue (Mike) :

- I'm desperately trying to get 夭, but always get 天, or 禾, or 头...etc. even though I draw the first stroke in the right direction, whereas 天 is recognised whatever the direction of the first (and second) stroke. Actually first horizontal stroke from left to right, second from right to left - and the opposite, as well as both from left work well, but if both are drawn from right to left, it doesn't.
Similar observation of the "non-critical first-stroke direction" for 两, 听 (etc..) : the direction of the first stroke doesn't matter, the systems recognises them.

I don't think the algorythm is actually so "strict", since the direction of the first stroke is absolutely needed for 和, whereas it is not at all for 天 (or for any other character that I know, and I have been using Pleco for a few years). This is what puzzled me, by the way : being used to the excellent "fuzzy logic" of the recognition system, I was unable to understand why, if my 和 was exactly shaped as it should, there was no way to get it, and why this occurred only for this specific character.
Hope this helps.

Hi ziggy

I politely disagree with what you just wrote about the importance of stroke order, and that the character written by Captain Planet wouldn’t easily be recognized. As to the 和: I tried writing it quickly a couple of times, and it always ended up being very similar to the one in the picture above. Why is that? Because of the stroke order. Quickly writing the character with the correct stroke order will eventually result in a very strict pattern. I suggest you check out http://www.cidianwang.com/shufa/, select 草书 and enter the character 和.
The handwriting recognition is actually very forgiving, but I assume it works by stroke order: It recognizes the first stroke as heng, shu, pie etc, then narrows down the results, then waits for the next stroke and again narrows down the list of possible characters. This works the same way the “write character by stroke selection” in Pleco, or in many mobile phone IMEs work. (Of course I could be wrong and the HWR works by character shape or completely different)

ziggy said:

I don't think I agree with you about the fact that a scribbling done with the right direction of strokes, but that doesn't look like the character is "better" [...]

Click to expand...

That's not what I meant. Of course, don't write like that if it's supposed to be read by a human being.

What I meant: you can write like that and still the character gets recognized (by the handwriting recognizer). This tells you what is important (for the handwriting recognizer) and what is not. As long as the number of strokes is right, and the first few are done in the right direction, it doesn't matter what the drawing looks like (for the handwriting recognizer). I wasn't trying to pass a judgment here, just telling you how to use the handwriting input efficiently.

ziggy said:

I beleive however that number (and nature: i.e. dots, horizontal line, vertical line..) of strokes are more important than the direction of drawing.

Click to expand...

ziggy said:

As long as a stroke is of the right length and position within the final shape of the character, only its length and position/orientation matter, not the direction of drawing. Printed characters have no "direction" of drawing at all and still are perfectly well understood.

Click to expand...

You can't really consider the two in isolation. Change the font to anything with "Kai" in the name, zoom it up, and you'll see the strokes are thicker in the beginning and thin out towards the end. The look of the stroke tells you the direction. The direction implies the look, especially when you're writing with a calligraphy brush.

This character is a calligrapher's favorite, and I think a good illustration of how the stroke direction and the resulting look are inseparably tied:

We are in "violent agreement"... we all agree that the direction and order of stroke are very important for a computer to recognise a character.

I appreciate tour indepth knowledge of Chinese writing, and I thank you for the comments.

The only thing is that in my humble opinion, it should not be more important than the orientation, length and number of strokes, because as it is now, a beginner may very well learn to write 和 as you did - properly in terms of strokes orientation, but very oddly looking, if I dare to say - ... and when using it to communicate with a Chinese human being, he would fail.
I would rather learn to write a 和 that looks like a 和, even with a wrong direction in one stroke, than learn to write a 和 with all strokes in right directions, but impossible to understand for a human being.

And I insist that based on the examples that I gave, the system is not consistent, since it tolerates wrong direction(s) in some characters but not in others like 和.

Re that last complaint - the Enhanced Handwriting Recognizer is considerably better in this regard, but the reason it's inconsistent has to do mostly with the vagaries of this sort of pattern matching algorithm; an incorrectly written 和 in the way you happen to write it is a closer statistical match for a bunch of other correctly written characters than it is for 和. Have you tried the demo of the enhanced recognizer to see if it does a better job with 和? (it'll show you the results from both recognizers on the same screen so you can compare)

Thank you Mike for your comment.
My last statement was not a real complaint, I am very happy with Pleco and highlighting a very minor (and single so far) recognition algorythm "inconcistency" is just for the sake of analysing and understanding a puzzling obervation. I understand very well that this is not exact maths calculation, but rather "artificial intelligence" computation...so, never perfect by nature !

I just tried the Enhanced plug-in and, yes, I can find 和 even with the wrong direction of the first stroke. I will of course buy it right away. I've just downloaded "pa2hwrnew-140606.zip" from "https://www.pleco.com/androidextfiles.html" to my computer (my phone is not safe to do purchases). Do I have to unzip it then copy/paste it somewhere on the phone ? And how do I pay, please (paypal or Visa), from my computer ?

Thanks!

You don't need to download anything from your computer if you already downloaded the file within our app, as it sounds like you must have if you were able to try the demo - just buy it (using the in-app $ button or our website store at pleco.com/store - we take Visa/MC/AmEx/Discover and PayPal) and it'll immediately activate.

Sorry to be out of topic here, but I just purchased the add-on (+ Learner's dictionnary + Text-to-speach female), and got a purchase number.
On the screen that I had after my purchase (through the computer) it said that if I accessed the same page over my device, it would be enough to click the code... but I've gone to the address on my Smartphone and I only have a list of add-ons to purchase, no code. What shoud I do now please ?

Enter the code manually in that case (via the in-app "Registration" screen) - sorry, we should clarify that text; a lot of people access the store from their phone so the link is a helpful shortcut for them but it doesn't do much good if you're on a computer.