Palm Kills Off Graffiti

Ed writes "PalmSource, the company that makes the Palm OS, has decided to stop using Graffiti for text input in all future versions of its operating system. Instead, it will switch to using a version of CIC's Jot recognition system, which will be called Graffiti 2. PalmSource was forced to make this move after losing a patent infringement lawsuit brought by Xerox. Jot is already used by the Pocket PC operating system. You can read more about it on Brighthand."

slashdotted server already

[Anonymous Coward]

By the Brighthand News Team
January 13th, 2003
When you think Palm, you think Graffiti. But Palm's long-standing association with its home-grown character recognition software is about to take a dramatic turn. PalmSource, the operating system subsidiary of Palm, Inc., announced today that future versions of Palm OS will not contain Graffiti. Rather, they will incorporate a modified version of Communication Intelligence Corporation's Jot handwriting recognition software, something it's calling Graffiti 2 powered by Jot.

The impetus for the switch appears to be legal rather than technical. In April 1997, Xerox sued Palm, claiming that Graffiti was essentially derived from its patented Unistrokes technology. Unistrokes, or "Unistrokes for Computerized Interpretation of Handwriting", as it is referred to in Xerox's 1997 patent, is a system of text-entry using single-stroke symbols for computerized recognition of handwritten text. However, it appeared Palm dodged a legal bullet when, in June 2000, a federal judge dismissed the case. But in late 2001, Xerox won a reversal in the U.S. Court of Appeals and the lawsuit was back on, and it's been hanging over Palm's head ever since.

CIC's Jot recognition software has long been found on competing handhelds running on the Pocket PC platform. As with Graffiti, its alphabet is based on block characters. However, unlike Graffiti, some characters require two rather than one stroke. Therefore, Jot characters more closely resemble common block letters than Graffiti characters. According to Marlene Somsak, Palm's VP of Communications, this will reduce the learning curve. "For new Palm users, Graffiti 2 powered by Jot is more intuitive and natural than Graffiti," Ms. Somsak told Brighthand.

Hints to Graffiti's demise began to surface last year, when Palm OS licensee Handspring said it was dropping Graffiti in favor of integrated thumb-type keyboards for its Treo organizers. And Palm itself announced in November that, for the first time, it was bundling Communication Intelligence Corporation's Jot handwriting recognition software with its upcoming Tungsten W handheld.

According to Lee Williams, VP of Engineering for PalmSource, the move to Graffiti 2 will allow Palm Platform licensees the choice of foregoing the silk-screened "hard" Graffiti area, since Jot can accept input from anywhere on a device's touchscreen.

According to Mr. Williams, Graffiti 2 powered by Jot will be a modified version of the current version of Jot found on CIC's website. It will be included in future releases of the Palm operating system, including the upcoming Palm OS 4.1.2 and Palm OS 5.2, and will be included in the Palm Developer's Kit (PDK) as part of a unified API.

How do I "Jot"?

[Plutor]

Check out CIC's web site for information [cic.com] on JOT, as well as a listing [cic.com] of the symbols.

Re:Well, that's it for Palm.

[EdFromBrighthand]

If you're going to make people think about a completely new user interface, they're going to think about migrating to PocketPC devices, as well.

But they aren't making a completely new user interface, all they did was change the character recognition software.

slow and non-standard

[g4dget]

Graffiti combined the worst features of custom strokes and regular writing: like custom strokes, it required training, and like regular writing, it was comparatively slow. In addition, it required most people to look at the handheld.

Palm should have used something like Jot from the start, or they should have copied Xerox's Unistrokes better.

Here [yorku.ca] is some Unistrokes performance data showing it to be the fastest of the bunch. There are papers comparing Graffiti and Unistrokes directly, and, again, Unistrokes comes out way ahead.

Jot Usability?

[webword]

I know that Graffiti had very reasonable usability [yorku.ca]: "After one minute studying the Graffiti reference chart, about 86% accuracy is attainable. Following five minutes of practice, accuracy improves to about 97%. Without further practice, users demonstrate total retention after a one-week lapse, with accuracy holding at around 97%."

How does the usability of Jot compare? Any ideas? Personally, if I am entering text, I like to use a thumb keyboard (e.g., Blackberry). One more thing, I guess that Jot 2.0 is available as shareware [palmflying.com]. It gets good ratings, but I haven't seen any "real" usability research.

QuickWrite was a good alternative

[manastungare]

Ken Perlin [nyu.edu] from NYU developed a great alternative, QuickWrite [nyu.edu] that, inspite of a steep learning curve, can be faster for experts to use. Palm should have adopted QuickWrite instead -- but perhaps they wanted to remain newbie-friendly.

Re:Wouldn't this be patenting the alphabet?

[gwernol]

I mean, it doesn't matter HOW you write the text, im sure some people write in uni-strokes as it is with a pen and paper without even knowing what it is.. How could Xerox patent a writing STYLE? Can I patent the way I make a capital P? Absurd!

RTFPA (patent application). The patent is for "A machine implemented method for interpreting handwritten text..." in other words it is the method for reading uni-strokes that is patented, not the Unistrokes themselves:

The patent [delphion.com]

Jot is cool.

[Ageless]

I used Jot on my Palm when I used a Palm because I hated Graffiti. Jot is *almost* normal handwriting. If you just write on the input box you will pretty much get it right. There are just a few letters that you have to learn.

I found Jot much easier to use and learn than Graffiti and when people used my Palm they thought it was a lot easier.

Re:Well, that's it for Palm.

[Anonymous Coward]

I migrated from a palm to an Ipaq, and wow, how can a 64mb device be slower and less efficient than an 8mb device. Just porting over my existing stuff left me fuller on the new device than I was on the old one, not to mention the huge power consumption difference. I use my Ipaq at work ONLY, it won't last a day without a charge, but the palm will go weeks before needing it. Not to mention if the power on the IPAQ goes, so does all YOUR BLOODY DATA :(

Re:slashdotted server already

[Keith Russell]

I think there was some hackery around the X character in Graffiti. If you invoke Graffiti Help, you'll see that the official way of drawing it is as you described. Two strokes, top-left to bottom-right, top-right to bottom-left. But scroll down in the help to the extended shift page. Look at the multiplication character's stroke. Top-right to bottom-left. But the preceding extended shift stroke is top-left to bottom-right. And the resulting character isn't a distinct multiplication symbol; it's a lower-case X! Sneaky, eh?

For the record, there is a single-stroke X gesture. Just keep the stylus down between the two strokes of the "official" X gesture. Think of it as a lower-case alpha, or a sideways shortcut gesture. Either way, I found it easier than the two-stroke X, which I always slopped into "lower-case I, [CR-LF]"

Jot User's Guide

[redtail1]

http://www.cic.com/support/faq/Pos/Jot/JotUsersGui de.html [cic.com]

If there's a space in the URL, that's Slashdot's fault.

I wouldn't worry too much...

[Anonymous Coward]

It should remain a very simply process of installing a 3rd party piece of software to add Grafiti support, similar to what users of the Handspring Treo do now.

strokes for cursive

[yerricde]

Quite a few printed letters we write every day require multiple strokes to write them cleanly and properly.

That may be true for printed letters, but among Latin lowercase cursive letters, the only ones that need more than one stroke are i (need the dot to distinguish ii from u) and t (crossed).

Other alternatives

[Fluffy the Cat]

(plug) Of course, there are other possible methods of input than just handwriting-style or on-screen keyboards. Dasher [cam.ac.uk] is usable on PocketPC (the reason for the lack of a Palm version is that up until relatively recently, Palms just didn't have good enough screens to let it be used without everything turning into a blurry mess. There'll probably be a Palm version before too long now)

Re:Why don't they fight it?

[ptomblin]

The patent was granted in 1997. It was applied for some years before then.

Re:Why don't they fight it?

[pstemari]

The earliest Pilots I can find reference to came out in 1996. Xerox filed for this patent October 26, 1995.

Re:Was Inkwell even considered?

[Ponty]

Newly developed? Inkwell has existed for a long, long time. It was Rosetta, the printed recognizer on the Newton before the Newton was killed in 1997.

Strangely, I don't see why Palm doesn't have prior art claims (those apply in patent cases, right?) Graffiti was released as an enhancement to supplement the initially bad HWR on the early Newtons. I think it was available in '93 or '94 (though I don't know for sure.) Strange.

prepare to be educated...

[Anonymous Coward]

No way, no how is this an attempt by Xerox "years after the fact" to quash Palm's Graffiti. Goldberg and Richardson at Xerox PARC first proposed this in their paper "Touch-typing with a stylus" at INTERCHI '93. Folks in the handwriting-recognition (HWR) field recognized at least as early as November 1994 that Graffiti was a commercialized application of the concept embodied in Xerox's Unistroke (check out this old newsgroup thread about it -- http://groups.google.com/groups?q=g:thl577785495d& dq=&hl=en&lr=&ie=UTF-8&oe=UTF-8&selm=ameyer-101194 1032270001%40rzuslip3-1.unizh.ch) so Palm cannot reasonably claim that they would have been unaware of Xerox's work waaayy long before their product was available on anything but the Newton (to make that claim, Plam would have to be claiming they were unaware of general-knowledge within their one-and-only field of HWR).

What's more, Xerox received their patent in January of 1997 and started the suit against USR (who owned Palm/Graffiti back then) just 3 months later in April of that year (http://news.com.com/2100-1023-279360.html?legacy= cnet). Since you can't sue for violation of a patent while your patent application is still being processed by the USPTO, it seems clear Xerox acted with as much haste to protect their idea as the US Government's red-tape allowed them to).

So get off your high horse and blame Palm for knowingly using Xerox's work for 8 years (and dragging the suit out for the last 5 years!) instead. Put it in a different perspective -- when Microsoft stole Stacker's work and then delayed Stacker's lawsuit until it no longer mattered what the suit decided, did you blame Stacker for trying to quash MS's self-compressing filesystems long after the fact?

Re:Xerox patent on UNISTROKES?

[John Whitley]

This sounds to me like another bogus patent. If something is very easy to re-invent independently, it shouldn't be patentable. I thought patents were supposed to be non-obvious.

Ob. Disclaimer: I am quite opposed to current patent law and its application in practice. That said, I feel the need to provide an alternative view to your knee-jerk rewriting of recent technological history.

Unistrokes was quite non-obvious when no one had actually done it yet. The entire field of handwriting recognition was relatively new. While many so-called "innovations" really do fail the non-obviousness test, this simply wasn't one of them. In this case, 20-20 hindsight blinds you to the novelty of the idea *before anyone had thought of it*. (Think about the design of the paperclip for a moment, if you don't get this.) Moreover, Xerox didn't just think of it, they researched the idea to show that their design actually made sense from an CHI perspective. Their work was quite innovative in that era's handwriting recognition research.

I have a hard time believing that the Graffiti devleopers didn't know about Unistrokes.
Published work on Unistrokes was readily available in conferences, journals, and online. Anyone doing even a minimal literature dive for handwriting recognition technology would have found the papers. Even if the work was independent and/or prior, there's at least some technical guilt for re-inventing a well-known solution. MS gets bashed for not-invented-here syndrome all the time -- why not Palm?

And remember -- half or more of the battle is in seeing past the now and into that first great idea. That this idea is amenable to a straightforward implementation is a *feature*, since it was aimed at low-power embedded devices.

You stipulate that Graffiti was "re-invent[ed] independently". Do you have any basis in fact for this statement, or are you just defending your vision of Palm as The Innocent Victim?

I mean, it's a little bit more complicated than using XOR to draw a cursor, but not that much.

You are clearly ready for the marketing department. "Oh, that's trivial! The engineering teams can do it in an hour or two, I'm sure!" The innovation here is NOT algorithmic, but rather in the design concept and how it dovetails with a user-interaction model and an efficient implementation.

Fitaly Tap Keyboard Better Anyway

[Sounder40]

Although I used Graffiti for years, I could never get much past 15 to 20 words a minute. Within a week of using the Fitaly Stamp, a sticker that overlays the Graffiti area, I was routinely doing 30 to 35 wpm. And the best part was the low error rate that I saw with Graffiti. Their site [fitaly.com] has several testimonials of speeds over 80 wpm.

The advantage is that the keyboard is designed to lessen pen (stylus) movement based on common words. It is highly customizable and supports international characters. shifting, special characters, etc.

I like it and it works for me. I won't miss Graffiti at all. Worth a look if you're interested in alternatives.

Re:Why don't they fight it?

[xanth]

The Xerox patent was GRANTED in 97. They _developed_ Unistrokes BEFORE the first Pilot came out. Also, the patent is not quite as trivial as "single stroke" letters. Because the strokes are single letters, they can be superimposed in the same space, without ambiguity in recognition. Furthermore, the direction of the stroke can be used to assist the recognition algorithm. These were the ideas that were patented.

Re:Jot Usability?

[Anonymous Coward]

Personal observation: It takes 0 seconds of learning to get nearly perfect text entry. It's just plain text. (Actually I have trouble with the k character sometimes, but that's it). "It just works" is a phrase that comes to mind. Not scientific by any means, but that's the way it was for me (I do not now nor have I ever used graphiti)

Palm and Graffiti older than Pilots

[Jonathan]

Yes, but oldtimers may remember that Graffiti is older than the Pilot. Palm used to make Graffiti for the early Newtons (in the 'Egg Freckles' days), back when Newtons were the *only* PDAs.