How's the handwriting recognition?
The first question people ask me about the Tablet PC is "How's the handwriting recognition?" Microsoft and the various system manufacturers like to downplay the handwriting recognition because they believe that one of the great things about Tablet PC is the fact that you can just leave what you've written into Journal in ink. In other words, why bother recognising it?
To some extent, they're right. But, depending on what it is you do, you may have to recognise text more often than you'd expect. For example, how often do you have to enter a Web address into your browser, an e-mail address into your e-mail client, scheduling information into your calendar, or data such as username and password into an HTML form?. At time like these you will have to wrestle with the system's handwriting recognition.
I've tested a lot of PC-based HR engines over the last decade.
TabletPC's HR is the best I've seen. I was amazed at how it recognised some of my worst chicken scratch. But it's far from perfect, which means that it may have trouble with something that to you and me looks simple. The HR engine mangled the phrase "test word" to read "Dr test wor" with a carriage return inserted after "Dr." (View Screen Image) The HR engine really had difficulty with URLs and e-mail addresses, two data types that I use so frequently that my HR problems caused a serious loss in productivity.
TabletPC's HR engine, like most HR engines, uses a dictionary to help it identify the most probable word. E-mail and Web addresses typically contain words that are not found in these dictionaries. As a result, I found myself doing less than a third as many e-mails as I could with a keyboard, and my Web browsing activities were seriously curtailed. This was unfortunate, because the idea of using a tablet to browse the Web is just so perfect. Imagine sitting on your couch, just tapping links on the display. But the URLs are impossible to enter.
With respect to the e-mail problems, Microsoft officials agreed that the keyboard might be better suited for certain tasks. However, working without a keyboard is one of the advertised benefits of the TabletPC. Microsoft uses Outlook's ability to accept ink (as said earlier, a feature that requires WordMail to be enabled) as an example of how you can write e-mails in ink without text recognition. If I need the keyboard to be productive with the addressing of my e-mails, I might as well use it for other text that goes into them.
Where's Webster when you need him?
These problems reveal where Microsoft could have gone so much further on the integration front to ameliorate any productivity problems that could crop up due to HR problems. For example, the dictionary that HR uses should be dynamically updating itself with all of the entries in Internet Explorer's history logs as well as all of the names and addresses in the Outlook's address book. In fact, any application that has an auto-complete feature (e-mail addresses, web addresses, HTML forms) should have its caches dynamically loaded into the dictionary. Had this been the case, the HR engine would have recognised www.cnn.com, a site that I go to several times a day. But, no matter how hard I tried, the HR engine kept having difficulty with my rendering of those characters and it almost never worked.
In cases where HR isn't working, the Tablet PC Input Panel (TIP) can be used to correct the text. It has three modes: handwriting recognition mode, keyboard mode, and voice recognition mode. The easiest way to correct misrecognised text is to use the TIP's on-screen keyboard and to tap the backspace and arrow keys as necessary to move around the mangled text and make corrections. The TIP lacks an easy way to access the dictionary. It's bad enough that Tablet PC doesn't incorporate often used character strings into the dictionary. It's worse that there isn't a quick and easy way to add them myself.
Oddly enough, there is a way to quickly modify the speech recognition dictionary. I discovered a way to train the speech recognition engine for specific words that I use frequently, I trained it to convert the spoken words "Dan Farber" into his e-mail address with the semi-colon included since that's the way that Outlook separates e-mail address from one another in the To: and CC: fields. After training the system, it worked perfectly every time.
After figuring out that maximum productivity may require a combination of speech and voice recognition, I began to wonder why there wasn't a single dictionary for both speech recognition and HR, where I could add and modify entries by showing the system how I represent them with ink and speech.
