Tuesday 18 August 2009

Predictive Text Attention Deficit Disorders: what a load of shiv


If anyone hasn't seen the Armstrong & Millar "History of Swearing with predictive text" then watch it (see the bottom of this post)! All the mumbo jumbo that's been discussed this week in the broadsheets (the Telegraph) and tabloids (Daily Mail) about predictive text making kids dumb, impulsive and brings on Attention Deficit Disorders makes we want to send text messages full of expletives!

But fortunately the IT press got wind of this and thankfully some sensible observations from the Register made me cool down. Also, some of the online Daily Mail readers who posted comments made some really valid points, namley Jeff a school teacher in London who says school kids use multitap and don't use "predictive text" because it does not actually predict (and therefore save keystrokes). He says:

I'm a teacher and I don't believe school kids use predictive text. I've asked classes over the years and kids don't use predictive text because its "rubbish"; kids say it does not predict or help them type less.

Teenagers use "multitap" i.e. to type "c" they type the abc key 3 times. The issue therefore is NOT predictive text but the keyboard type. Kids are constrained by the keyboard's physical size.

So either, tell the phone manufacturers like Nokia, Samsung etc to only provide kids with phones which have qwerty keyboards (not likely to happen) or put in better predictive text.

Most phones use T9 which is old. Personally I use Adaptxt software on my Nokia (downloaded for free from the Adaptxt.com website) which does actually predict in advance and does encourage me to use proper English. Some of the pupils in my class think its rather good and I'm finally cool (not book)!

One final comment on this research: what a waste of time. Their initial assumption was wrong!


To delve a little deeper then: existing predictive text solutions (so for the majority of the world that is T9) just try to complete words and even then it doesn't do a very good job of that; T9 doesn't learn from user behaviour so makes the same stupid word suggestion every time. In other words, as user experience goes, "predictive text" sucks.

But as Jeff the teacher from London points out, what are we going to do about it? He rightly says, if the world wants kids to stop using predictive text and compose messages in their full language as opposed to phonetic versions, then handset OEMs need to supply mobile phones to teenagers with qwerty keyboards. Currently, qwerty devices target business users and if you review any telco or OEM device portfolio for teenage devices, very few are qwerty in form factor. So maybe there is no demand. If there was, surely OEMs would cater to this message-friendly user segment? Fjord, now home to Christian Lindholm (of Nokia fame), say qwerty for teenagers will be a growing trend as these young social communicators want efficient input.

I don't think the driver for this change is going to be concerns over Attention Deficit Disorder but 2 things need to be realised:

  1. predictive text as we know it isn't actually predicting. Rename it. "Word completion" isn't catchy but that is all it does. There is no prediction so school kids (and let's face it lots of other people too) won't use predictive text because the experience does not match the expectation. Yet, there are a number of input technologies that aren't being used: Jeff mentions Adaptxt but there is also Cootek, Wordlogic and others that should be considered. Has anybody actually compared these input technoliogies? Perhaps better "predictive text" will increase the number of emails a user sends or the number of times they post to Facebook from their mobile?
  2. the industry needs to educate users that qwerty hard keyboards are the most efficient method of input. 12 key maybe de rigeur in Western Europe and Asia for non-smartphone/text-heavy users but text entry on a qwerty will make for a superior experience.
Decision makers responsible for input technologies at handset manufacturers need to examine input user experience: Fjord rightly point out that Twitter is going mainstream, Facebook updates and wall comments are exploding. So handset manufacturers need to give users devices which allow them to express themselves fully.

New input technolgies can assist the user's messaging experience. Qwerty handsets for young social communicators will encourage mobile blogging, mobile Facebooking and Twitter updates. And does all this not drive demand for mobile internet tarifs at telco operators?

This is a win win for the industry and users. And as Armstrong & Millar's Predictive Text Commission would say, that is not a load of shiv.

Tuesday 21 July 2009

I hate app stores. We ISVs need to get smarter than the industry followers of fashion!

The BBC Technology site ran an article yesterday "Apps to be as big as the internet" which sounds great but underneath the bonnet is a very scary forecast of the future for mobile app developers: basically, we're all like pop bands vying to be Top of the Pops. As Laurs from Getjar and as Sebastien de Halleux of Playfish confirms from experience, ISVs in the mobile space are one hit wonders. How on earth can ISVs run a sustainable business this way with such high but shortlived peaks and troughs which are long in time and deep in cash burn!?

Personally I don't know if I am a fan of the new Symbian brand identity and I also wonder whether their PR machine is needing some oil: an article on the BBC site 4 days earlier announcing Horizon aka Symbian mobile app catalogue is coming and yet in yesterday's newspiece, Lee Williams aparently says the industry doesn't need anymore app stores and the concept is flawed. Nevertheless, despite dodgy branding and slightly misaligned PR from Symbian, I agree 100% with Mr Symbian's sentiment.


Dealing with Dichotomy in the ISV business model
Ok, there is the app store. Not a business model any ISV wants to rely on to grow a sustainable business.
The opposing view for ISVs? In a previous post I referenced Vision Mobile's 100m club i.e. the ISVs who have their wares embedded on over 100m devices. "Embedded" is the important word here folks. Get embedded and you get revenue. How good the revenue is of course depends on how highly valued your software is perceived by the OEM and on your negotiating skills when it comes to Request for Pricing docs and beyond. The problem is it is hard work getting embedded. Putting aside levels of funding and business processes you'd need to have to become a trusted supplier to a global OEM, you also need to get real with your embed price. And OEMs will drive it only one way! So maybe embedding as a business model for ISVs stinks too?

Life enriching apps vs Hedonistic apps
The ISV needs to take a look at its products and decide whether they are just app store fodder or the apps developed will "enrich your life" as Lee Williams says. Hedonistic apps won't be embedded but apps that are either core to driving the user's personal communication network (solutions and technologies enabling messaging, voice, sharing, location, context) or driving media consumption have every chance of being embedded. If apps drive data usage, you're on the right track. Just get some good product marketing to align your offering to the OEMs handset strategy and sales people to get the right meetings at the right OEMs or MNOs.

Embed+Modular Architecture fit into the handset software lifecycle
Embed is the business model every ISV worth their salt needs to follow. But ISVs need to get smart on maximising the value from their product. OEMs are driving down BoM costs so get modular, split your product apart. Get money for embedding even if it is low BUT BUT BUT have hooks into your other modules which are your products or services users can find and will pay for post sale (see Handset Software Lifecycle by Vision Mobile). Share revenue with the OEM and the telco operator. This isn't easy but its better than drowning in an app store among 1000s of other sinking apps where no one knows your name.

Lee Williams says "give the consumer the opportunity to wander down a really relevant aisle of content and applications that they can get access to" so don't rely solely on an app store. Get embedded and take users into your own "store" of highly relevent products and services which plug into your embedded app. If you implement this well, the user doesn't really think "I'm shopping for apps". They are using the functionality anyway and a micorpayment for deeper features that enrich the experience is a no brainer.

Sat Nav is a great example. Nokia maps is useful to a point. The user can find their location and get routes but if you want turn by turn voice directions or city guides, you need to pay a few euros. The Map app has its own little ecosystem of software and services the user can quickly purchase.

Last.fm could be applied to handset in a similar way. Mobbler is a symbian client for Last.fm and stream music to the device and provides links to Amazon for download sales. Why is a handset OEM not partnering with Last.fm? its social, it drives data usage and there is revenue from downloads.

KeyPoint Technologies is reinventing predictive text. Yes, they embed their Adaptxt software but let the users buy additional languages for a few euros not stored in Rom and share revenue with the OEM. This removes the painful inflexibility of language packs for the user and the OEM sees revenue from text entry. Input is central to a users device experience. What other apps and services can be plugged into predictive text?

Users hate searching but love downloading
I don't have any empirical evidence but I believe Apple's app store doesn't prove users love hunting around catalogues. It proves they love downloading direct to their mobile. Get your AddOns or plugins intelligently positioned in front of a user via your embed, and the downloads will follow.

And look, the blog trend analysis tool says the same. Downloading is king. App stores ain't that exciting!

Wednesday 1 July 2009

What's wrong with using a virtual keyboard? Something according to Swype

You may have heard about Swype recently. It’s from the makers of T9 text input. Well, they’re on the PR offensive again.

In case you’re not sure what Swype is, here is some copy from their latest press release entitled “The Quest to Make a Better Virtual Keyboard”:

“The users holds down their finger and drag it from one letter to the next until they’ve spelled out a word. It’s surprisingly forgiving, so if you just barely miss one of the letters, it will probably still recognize it.”


(you’re guess at what’s being Swyped is as good as mine. Swype’s guess will be better apparently)


What’s wrong with using a virtual keyboard the way it is? A full key (or QWERTY) keyboard has the majority of characters you need available to quickly bash out an email or SMS (and if you have predictive text switched on it’s really quick!). A full keyboard is also an item that many people have grown used to, whether it’s a keyboard at work, on their iPhone, windows mobile or Android device. Swypes legacy, T9, was a good product because it changed they way people used their phones for messaging over a decade ago. It’s now an iconic product. (Love it or hate it) It evolved from a need for better text entry on 12 key devices. I’d be reluctant to say there is a need to change the way we use a virtual full keyboard though.


Using Swype requires the user to learn a new way of inputting text. Why? What’s the benefit? Swype guesses what I’ve typed after I’ve moved my finger (or stylus by the looks of the screenshots) around like I’m chasing an ant on the screen. I don’t want my predictive text to tell me what I’ve typed after I’ve “Swyped” it. It’s not predictive text then is it? And if it doesn’t predict what I’m typing, how am I saving keystrokes, or typing faster? It doesn’t seem as if I would be. I’d still have to Swype over every letter. What if it gets it wrong? I’ll still have to manually input the word. Wouldn’t I be better just typing?


Swpye’s objective is to “[be] the default text input on all devices, ranging from phones to tablets, TVs and other surfaces”. I’m not convinced. The user faces a steep learning curve learning how to swipe including a range of gestures that steer away from what you’re possibly used to if, for example, you use a windows mobile device. And it doesn’t appear to add any value to the user.


So, will it make me text quicker and more efficiently or improve my experience?


I doubt it.

Friday 26 June 2009

Is LG really innovating when it agrees to use T9 and xT9 for text entry on its future mobile phones?

Well, that's the view of Michael Thompson, senior vice president and general manager, Nuance Mobile. Don't get me wrong: the guys who have been with T9 pre Nuance have done an amazing job. Looking at Vision Mobile's latest release of the 100million club, who wouldn't want to say they were part of that. And hey, the Nuance acquisition of Tegic (and of course Zi earlier this year) firmly positions Nuance as the 800 pound gorilla in the text input market. But, does it mean innovation for OEMs and end users??

I'm not convinced. Nuance have been on an acquisition spree over a number of years, grabbing talent, contracts, channels and customers for their assortment of products and services which make up their mobile division. Seeing that lot integrated is one heck of a job. Seeing them deliver truly innovative solutions over the coming number of years is another. Does Nuance have a culture of innovation? Well, I don't know is the honest answer.

But what I can say is that from the perspective of an OEM, it can't be good news. SMS is still the killer app (see my previous post). If OEMs are using the same old input method across all devices, where is the differentiation? Cootek, Adaptxt, Tengo and other ISVs are in the text input game and have many fans.

LG is playing it safe.

Thursday 25 June 2009

Adaptxt for Nokia S60 Smartphone (Beta v2.4)

Following the release of Adaptxt version 2.3 in February 2009, I awaited in anticipation for the latest release of Adaptxt and its intelligent predictive text software. I loved version 2.3 but the application did have one or two irritating bugs, that I hoped the future update would erase. To my delight version 2.4 arrived much earlier than first anticipated (April) and didn’t disappoint.

The experience
The download process was very self explanatory with an assortment of pop-up messages to smoothly guide you through the installation process. It took around 3 min’s to download completely. This version seemed to be around 2MB in size, so I installed two additional languages (Spanish & French) and the application ran as smooth as a baby’s bottom.

I noticed that the personal dictionary default size had been increased from 500 words to the new default setting of 2500 and that instantly the suggestions where more accurate. Also, they seem to have increased the size of their already substantial selection of languages. So if you’re multilingual you’ll be sure to find an array of language options to choose from.

New features (v2.4)
This new version introduced PIM & SOCIAL into the functionality of the software, allowing you to extract your social networking profile and other personal content from your phone into your personal dictionary.









PIM
Allows you to choose what Adaptxt should scan, so you have the choice of scanning emails, SMS, domains, contacts and calendar. So basically Adaptxt learns all your files you write and read.

Social
This application allows you to connect to Facebook (the only social networking service at present) where Adaptxt will pull your profile information to learn new contexts for your personal dictionary.

Finally…
On a whole this release was worth the wait. Adaptxt version 2.4 definitely is more intelligent in that it’s more accurate in suggesting word and phrase suggestions in advance, than its predecessor. It’s certainly still my preferred text input software by far.




Qwerty keyboards rule for SMS. Watch out touchscreens!

I know everyone loves the iphone but as Tomi T Ahonen reminds us in his epic essay (but is well worth the read), iphone is 1% of the global mobile market and its not SMS friendly. Is he right? Well, yes.

I was at the Nokia conference last December in Barcelona when a VP with Finnish name (not surprisingly) announced the N97. During the course of a good launch presentation, he wowed the audience with touchscreen controls, widgets and all Nokia's wonderful new mobile services. BUT what he kept right to the end was the qwerty keyboard. I remember his saying (and I paraphrase here) "you cannot beat a qwerty keyboard for messaging" and tadaaaaaaaaa the giant N97 on the giant screen behind him started to split before our eyes and yes, a simple qwerty keyboard emerged! (some from Slashphone has kindly uploaded images from that presentation)

Tomi Ahonen's article is a historical perspective on the importance of messaging, the role it places on keyboard design and the resulting success of a device i.e. people like phones which are messaging friendly. He points out Nokia are the kings of SMS and Samsung have worked hard to follow suit. Check the link to the leaked Samsung roadmap showing the T559: yes it looks guff and yes its not for me (device target demographic is probably teen or student girls) but its a messaging device....you know for these young "social communicators"!

Thank the lord. Touchscreen's look good but for SMS messaging (the killer app on a mobile) touchscreen's suck. Oh, and using a stylus ain't cool either!

Tuesday 23 June 2009

Android Soft Keyboard & Predictive Text

I waited a long time for the Cupcake update. As well as improving performance, UI tweeks and adding video playback, it introduced a soft keyboard (SKB). No longer would I have to flick open my G1 to email, text or twitter! Word completion and predictive text would make my life easier again!


Did it?


A few weeks on and, to be honest, I’m only just starting to use the SKB more. Like the iPhone, the G1 SKB in portrait takes a bit to get used to and can be tight. The accuracy of keyboard is very good, however it’s the accuracy of my thumbs that’s in doubt. This makes typing slow. The proximity error correction is ok and does help but it doesn’t help enough to stop me from bashing away on the hard keyboard instead.


Auto screen rotation is the Android SKBs saving grace. The keyboard instantly becomes thumb friendly. It takes up just over half the screen but doesn’t look cramped. Even when the predictive text kicks in, there’s still enough room up top to see what you’ve written for a few lines. Additionally, you can switch on haptic feedback. It’s a love it or hate it feature but personally I love it. It’s nice to “feel” a soft key being pressed.



The best thing I like about using the SKB is the predictive text input - my spelling safety net. It’s not the best I’ve seen but for a person who has used predictive text most of their mobile life, it was good to get it back after typing blind for a few months. Suggestions don’t appear till you’ve typed a couple of letters, then a bar pops up on top of the keyboard. Suggestions are generally ok. I guess they’re using the same statistical model that T9 operates with. For longer words it’s easier to type the full word than mess about scrolling right trying to find the word you started typing with it ending in “ed” or “ing”, for example. At best, I think the predictive text is only saving me from typing a few keystrokes here and there.


The SKB hasn’t quite made my life easier. I still flick open the keyboard, particularly if I’m typing a long email or text. Don’t get me wrong, the SKB is great and works well but I think that in wanting a SKB for so long has made me realise, once I had one, that I wouldn’t trade in good old hard keys for anything else.