Voice recognition systems have come a long way in the past five years. Now, surveyor Ridge and Partners is using the technology to stay one step ahead of the competition.
The future of Ridge and Partners is in the hands of David Walker. He is responsible for research and development at the multidisciplinary outfit, which means that he has the job of keeping his firm at the forefront of the industry. This is a demanding enough task at the best of times – and these are not the best of times to be a medium-sized QS. As smaller firms are swallowed up by major players, the survival of those that remain hinges on their ability to come up with radical ideas to improve competitiveness, which is where Walker comes in.

His latest brainwave is to introduce voice recognition software. This will allow PC users to dictate directly to their computers, rather than to a secretary. If it works, it will dramatically cut the two days it often takes a building surveyor to get a site report to a customer, not to mention the £12 000 a year for each temporary secretary .

Walker first became interested in voice recognition technology five years ago. A neighbouring firm of solicitors in Oxford was trialling software that would allow the lawyers to dictate reports, letters and memos directly to their PCs. Walker managed to wangle an invite to see how the system was working.

His expectations were dashed as he watched one of the firm’s partners talking to his PC like an automaton. “Dear-Sir-comma-new-paragraph,” the lawyer chanted in a monotone staccato. “I-am-writing-to-inform-you … “

A disenchanted Walker decided to give software houses a chance to refine their product before investigating further.

Improving with age

Five years down the line, and after an impressive demonstration by software house Dragon, Walker is trying out voice recognition with his building surveyors. The trial started three months ago, when the software was installed in Ridge’s Oxford office, where it eventually ended up on the PC of Paul Fenton, building surveyor and, now, guinea pig.

Fenton was pessimistic at first, but after a cautious start using the software to write letters that would usually be dictated to a secretary, as well as faxes and e-mails, he is a convert.

Fenton demonstrates how it works by raising his mouse-cum-microphone to his mouth and saying: ”If you should have any comments or queries.” The words “should you have commercial queries” appear on the screen. Fenton corrects the text by repeating the words and continues: “The original signed copy will follow in the post,” which is interpreted as: “The regional signed copy will follow in the post.”

How to train a PC

Despite the misunderstandings, the demonstration is impressive. It is faster than giving a dictation to a secretary, and much faster than Fenton using two fingers to bash out the document himself.

The system takes time to get to know its master’s voice. Users log on, create their own voice recognition file, and then train the PC to recognise their speech. Software consultant VoiceWrite recommends starting with short documents before working up to large reports.

Walker himself is from Lancashire and had particular trouble persuading his computer to understand his pronunciation of the command “done”. In fact, it took a long process of trial and error before the PC learned to transcribe his flat northern vowels.

The other glitch is specialist vocabulary. In training sessions, users stop at every technical term to type in the spelling and teach the PC to understand it. This usually has to be repeated three times before the PC automatically recognises the new word. The routine applies to regional accents, too. But Fenton has been amazed at words the system does know; so far, “tenalised” has been the most obscure.

Overall, Fenton is pleased with his progress. “I have found it recognises my voice quite quickly,” he says. “I have got as far as dictating minutes and faxes.” But he admits there are times when he has to go back over text to correct the computer’s mishearings: “It makes mistakes, but in two to three weeks I have seen improvements.”

Other problems have included finding a suitable position for the trial computer. Walker explains: “We set it up on an empty desk at first, but there was a reluctance among the surveyors to hot desk because they weren’t next to their phones.

“We then put it in the centre of the office but we found that colleagues would pass the desk and say ‘hello’ in the morning, and ‘hello’ would appear on the text.” The machine is now on Fenton’s desk, but those around him have to keep their voices down when he is talking to it.

The trial has cost Walker £2500 in hardware and software. Dragon charges £250 for a half-day training session, but the major cost is the amount of time the building surveyors need to spend getting to grips with the system and training the PC to recognise their individual voices.

Voice recognition software goes on site

The software faces its greatest test next month, when Walker has scheduled Fenton to take it on site. The problems Walker found with extraneous noise and accidental interruptions in the office will be magnified, as will the disaster if any stray comments find their way into the final document – Walker admits that one of his main worries is the colourful language used by contractors. Unfortunately, ”the system does recognise swear words”, he jokes.

Walker has earmarked £20 000 for introducing voice recognition software to the firm’s 38 building surveyors, who are divided between four offices in Oxford, Bristol, Reading and London. He calculates that this will eliminate the need for temporary clerical support – the Oxford office alone has enough typing to keep one temp fully employed. Voice recognition will also leave the permanent secretaries free to concentrate on more demanding administrative duties.

But, essentially, the benefits from voice recognition will come from upgrading the service Ridge can offer to clients – above all, a quicker turnaround on site reports. In the battle for survival in the quantity surveying market, Walker is hoping that the software will give his company the competitive edge that could make all the difference.

IT to shout about?

  • Reduces the need for secretarial support
  • Cuts down turnaround time for building surveyors’ reports
  • It takes time to train the software to recognise the user’s voice
  • Surrounding noise can interfere with the dictation