All my own Werdz

This is the text of article that appeared on Page 1 of DEVELOPMENTS at Hursley, Issue 8 - 30 April 1999, produced by Communications and External Programmes for the people of IBM Hursley.

Just before my son was born, in November 1985, I gave a paper to the Acoustical Society of America in Nashville. The paper was entitled ‘Aligning phonemes with their corresponding orthography’.  This simply means determining which letters in a word are spoken with which sounds.  The development of this algorithm had enabled me to perform validation checks on the pronunciation of words I was generating whilst building a dictionary of pronunciation for use by the Speech Synthesis group at IBM’s Scientific Centre in Winchester.  I completed my assignment and returned to Hursley Park but I was one point short of my first Invention Plateau which I thought I could get by writing a paper on spelling ‘sossidge’ correctly.

Time moved on.  My son Stephen grew up, went to school, became proficient in maths and learnt to use computers.  At one
parent/teachers meeting the maths master suggested I encourage Stephen to learn to program computers.  We’d bought an Aptiva when he was ten, so all I had to do was find a good programming language for a ten-year-old.  I asked around and bought Delphi 2.0 which I gave to Stephen for a Christmas present - what a wonderful father!  My son would have preferred ‘Total Annihilation’ or some similar ‘shoot-em-up’ computer game.  Needless to say any attempt by me to teach him to program was diverting him from playing games.  Stephen did eventually produce a program, though, which helped him learn a long list of Latin words.

However, as a person who works for IBM, I’m often asked if  I could teach someone how to use computers.  I accepted one particular challenge and went to their house to meet the computer.  It was an ancient machine, almost ten years old (!), but it did have a diskette drive.  A thought occurred to me.  Somewhere I had a set of floppies with an abridged version of the dictionary of  pronunciation I had built.  Could I copy them onto a set of diskettes, write some code using Delphi, and correctly spell ‘sossidge’?   I asked if I could copy the floppies but when I found them and thought I had finished copying them I discovered one was missing.  I explained my problem to my wife and she said she’d made a copy of the floppies years before but did not know where they were.  After much searching and accusations of untidiness we found them.

In December 1997 I started the project.  I had an operation on my shoulder in January 1998 and was off work for a fortnight, so I could devote quite a lot of time to the project.  My aim was to use Delphi to produce a tool with a simple user interface which could be used to spell-check a word that had been spelled as it would be spoken.  I queued up to use the Aptiva.  Like Stephen, my wife Jane is also afflicted with the desire to spend hours playing computer games, so the only time I was allowed to use the computer was either late at night, early in the morning or on a Saturday when it was raining.  We had a lot of Saturdays like that last year.

By June I’d got something I could show to colleagues.  I’d also thought of a name for my project - Werdz.  Those to
whom I showed it encouraged me to continue its development.  My family was bored with it but were sympathetic, as others said it was worth pursuing.  But what does one do with a piece of software that is not closely related to one's current project at work but should be of use to someone somewhere in IBM?

I spoke to Intellectual Property and was reminded of the agreement I’d made with IBM that inventions and copyright works made by me in IBM's field of interest belong to IBM.  In August I submitted an invention disclosure.  In October a patent application was filed, so I did'nt just get one point for a publication but three points for a patent application, a file award and the Invention Plateau.

However, another thing also happened in October.  The WebAhead conference took place at Hursley and I was the chair of the session that included a presentation on alphaWorks.   At the end of the session I spoke to the speaker, Chris Bahr, and asked if my Werdz project would be considered a suitable submission.

I made my submission and in November started corresponding with John Wolpert, a member of the alphaWorks team.  For Werdz to be accepted, it had to use an IBM-owned dictionary of pronunciation.  It would also help to market it if the definition of a word could also be displayed.

I got permission to use the dictionary of pronunciation used by ViaVoice and adapted my code to use the IBM dictionary.  When I started this project I had purchased a copy of Collins English Dictionary, but the documentation that came with it did not say how one could use one’s own code to display a word’s definition.  I emailed the Collins sales desk to ask how I could achieve what I wanted to do.  Collins were very helpful.  After a few iterations I got the instructions I needed and proceeded to develop an interface from Werdz to a dictionary.  I also developed an interface to access Werdz from another application such as Notes or WordPro.  I reported back to John and he put into motion the process of publishing Werdz on the alphaWorks Web site.

On 2 April, 15 months after first beginning the project and with 1,000 hours spent on development, Werdz was published by IBM for evaluation at:

Queuing for our Aptiva is no longer a problem.  I’ve spent most of this year at a customer site in Belgium;  Stephen was rewarded with his own computer for gaining an academic scholarship to King Edward’s, Southampton;  so that leaves lots of time for Jane to play on it.

I have had a great deal of enjoyment and satisfaction from this personal project and I am most grateful for the encouragement I have received from my family and colleagues.  But I dare not tell my family about my next project yet!

Werdz spelz oaquaigh