The Pros and Cons of Semacode and Other Visual Tags (UPDATED)

Update: Based on the comment that Stan made, I retried the experiment and got it working. Viewing Distance was the trick. Check the comments.

Someone slipped me an email on semacode, so I decided to look into it. It struck me as interesting and worth taking a serious look at. After all, mobile technology and groups such as MobileActive (which I am a part of) use mobile technology for all sorts of things.

On the surface, it's interesting though I must say that a world full of images (as to the right) does not strike me as something I would consider attractive. When I read about it, I thought, 'cool'. When I saw a semacode, I thought, 'Wow, that is pretty ugly'. Coming from me, a very forgiving critic when it comes to visual art (I cannot draw a stick figure without concentrating), that's pretty damning.

Still, it's interesting and useful for a wide variety of things.

Taking It Through The Paces

So I decided to try it out. I headed over to Semacode.org and decided to make a tag instead of reading what is probably marketing brochure material. The proof, they say, is in the pudding. So - on to the Semacode tag generator.

I created a tag for 'www.knowprose.com' (which you can see to the left) and found that this required Java, which I am not fond of, and also required me to do a screen capture and edit to get the image down to size. What a travesty that they couldn't generate the image file there themselves.

So, now I have a tag. KnowProSE.com is semacoded. Somewhere, someone is really impressed. Meanwhile, I had to download the software for my Nokia 6620, which only works in Microsoft Windows¹. Clearly, I wasn't prepared for this experiment on this PC, and I don't feel like firing up the laptop right now. Change of direction.

Semapedia

While that's happening, I continue reading and find Semapedia, which codes entries to the Wikipedia. OK. 'Taran Rampersad is in the Wikipedia. Sure enough, I fed it the tags and it generated tags for me (attached, below)². OK. I printed the tags out, downloaded the Semapedia Reader to my phone, and installed it. I ran it, and it brought up the camera for the Nokia 6620 - where I snapped the image in ambient indoor light.

It couldn't decode it. Oops. I tried it from the screen. It couldn't decode it. Oops.

Sorry, Semapedia, you got two tries and they both failed. Now, it could be a connectivity issue of my phone here in Trinidad and Tobago - through Telecommunications Services of Trinidad and Tobago, or maybe my camera's images were too difficult to translate, but... at the end of the day, it didn't work and told me it couldn't decode the Semacode. That error message implies that the semacode doesn't work, at least for me. However, it did work for the default Wikipedia front page.

Nicely done for an existing example, but my example didn't work. That sucks.

Perhaps it takes time to register in the database? I'll try later and comment whether it did or did not.

Of course, that may be user error somewhere too. I checked the Semapedia Wiki FAQ, and there wasn't anything there. Maybe my phone just didn't know the secret handshake. Maybe it timed out. Or maybe it doesn't work. I don't know. But the idea is... OK. If it had worked for my example, I'd be much happier since I could have explored it better.

However, I didn't see languages other than English noted - and it seems to me that the power of a symbol such as a semacode could be used to transcend language - that a person should be able to select the language of the Wikipedia entry in the reader itself. This isn't presently done, but it could be quite useful. And then, imagine - for translations in foreign countries! Brilliant. But as someone else pointed out in the FAQ, maybe it's too complicated. Maybe OCR would be better, though the code overhead for OCR might be a real issue.

Imagine

When I was a teenager and 640 kilobytes of memory was considered a lot, I showed my late Uncle Amar a sketch of an idea for object recognition - it was back in 1988, the year I left Trinidad. I had labored over this idea, taking pains to draw it all out. Basically, the system would rotate an object and take snapshots of it, and based on what it was told the object was, it could identify that object from any angle. I never did follow up on that, and the haze of code and other things over the last 18 years (!) didn't allow me to try it out. I'm sure someone else had thought of it before, or after, and did something similar. Still, the technology is getting there... and while Semacode, and QR Code, and things like them are interesting - they are just barcodes. And if the images have to be perfect to translate, one has to wonder whether the tags themselves will present a problem if they are not 100%.

So there are problems.

But then, imagine systems that can recognize objects from 2 dimensional images. It could be done, and maybe the Semapedia isn't perfect, but the romantic idealist named Taran likes the idea that this can happen, and also likes the idea that even if the Semapedia and other things do not work perfectly and the tags are ugly... it is a step in the right direction. Personally, I think creating tags for objects is a bit silly when we could be identifying the objects themselves.

Should the semacode system work, though, it could be used for a lot of things. I've been consulted on for all sorts of inventory related issues, and semacode could be used for it - and be cost effective, since it doesn't require special equipment. In one sense, it could be good for businesses. In a slightly adapted direction, it could be used for helping people decipher things from landmarks to directions to a restaraunt. A lot is possible. Think of a fire hydrant which could be checked by an inspector periodically, and anyone could see when it was last inspected. Imagine going to a restaraunt and pulling up a list of reviews.

Imagine.

And yet, I keep wondering why we worry about tagging things when we could just identify the images themselves. A world with semacode tags on everything seems unappealing. There has to be a better way, and I think the key is object recognition... not 2 dimensional black and white tags. But to get there, we have to be here, and it's glad to see that here has advanced a bit. Or a byte.

¹And, I might add, it doesn't have a way to turn the Nokia Suite off unless you go killing processes. Nasty code, Nokia, nasty code. Resource waster.
² If someone tries to stick one of these on me and they are not a scantily clad disease free woman that I find attractive, the response will be as violent as I feel... ;-) That may sound like a threat, but it's actually an invitation for the right people...