Matthew Cornell

The problem: Data is still frequently transmitted as semi-structured ASCII text, but applications require structured information. For example, I often want to input an address into an electronic address book based on text from an email or web page, esp. a signature. (Yes, there's vCard, but I think consistent usage is still rare, and using it requires extra steps.) Another example is inputting financial transactions into a finance program from an on-line banking page - I now do much of my bill paying on-line using a web interface from my bank, and I hate having to laboriously (and erroneously) copy each section of the transaction's output (HTML) into GNUCash.

The solution I'm playing with has two parts: a) a general-purpose data recognizer clipboard that converts semi-structured text to XML, and b) support by applications for recognizing XML on the clipboard. The thought is that by converting to XML (i.e., semi-structured data that has been marked up), we've done the hard low-level data recognition work, leaving it to the applications to 'take what they can use'.

Usage scenario: consider copying the following email signature (from this example) to the clipboard:

    Jim Frazier, President
    The Gadwall Group, Ltd. -  IT and Ebusiness Strategies
    Batavia, Illinois     630-406-5861      jfrazier@gadwall.com
    http://www.gadwall.com      http://www.cynicalcio.com
    Seminars and Training - Consulting - Publications

(I don't know him - it's just the first public signature I found.) It's easy to envision a straight-forward regular expression-based tool that pulls out the following:

    <clip>
      <name>
        <first>Jim</first>
        <last>Frazier</last>
        <title>President</title>
      </name>
      <location>
        <town>Batavia</town>
        <state>Batavia</state>
        <country>Batavia</country>
      </location>
      <phone>630-406-5861</phone>
      <email>jfrazier@gadwall.com</email>
      <url>http://www.cynicalcio.com</url>
    </clip>

Here's what the text copied from a bank's on-line statement might look like:

  04/27/05 | Checking | Check 2067 | 2067 | $-45.00

Where the columns are: Date, Account, Description, Check #, and Amount. In this case the date might be:

    <clip>
      <date>
        <month>04</month>
        <day>27</day>
        <year>2005</year>
      </date>
      <text>Checking</text>
      <text>Check 2067</text>
      <number type="integer">2067</number>
      <currency unit="dollars">-45.00</currency>
    </clip>

You get the point - basically it's just a set of lower-level recognized data. It's up to the application to put the pieces together in a more specific and meaningful way.

Thoughts:

• It would be great to program custom rules using a nice scripting language, such as in JavaScript (used to program Konfabulator widgets).
  
• As a work-around to requiring applications to recognize XML, we might try the XML clipboard plus a mouse/keyboard macro playing program (e.g., QuicKeys).
  
• There would need to be a DTD standard for this. (I'm sure one exists somewhere.)
  
• Hasn't this been done (partly) by Apple's old Data Recognizers idea? A few references here and here.