This talking shop gets results

NEXT time you phone your bank to request a new chequebook, check your balance or ask where their nearest cash machine is, you could end up having a detailed conversation without realising you’re talking to a computer.

That’s the vision of Marc Moens, chief executive of Edinburgh-based software firm Rhetorical, which has developed technology that turns text into natural-sounding speech.

The company has just celebrated three years in business by reporting a 70 per cent surge in turnover to 1.7 million, with gross profits more than doubling to 1.1m for the year to the end of January.

Mr Moens says Rhetorical’s success is down to rVoice, the firm’s innovative software package which converts the typed word into speech that is almost indistinguishable from the human voice.

"It hasn’t been easy - it’s a tough market out there," he says. "But we have a good product and certainly have the right strategy because we lined up some really good customers last year and continue to do so."

Clients include mobile phone operator Orange, the New Zealand Meteorological Service and Verizon Wireless, the United States mobile phone giant.

Mr Moens says: "Our first product was English - both Scottish and southern English - before moving into American and Australian voices, because when we started out most of our customers were in the English-speaking market.

"Landing Verizon Wireless was a very nice deal for us. They’re using it for e-mail reading, and everyone seems to be interested in offering that to their customers. Ananova, the news website, uses our voices for reading the news on its website, while Orange mobile phone users can listen to news over their phone."

Under the deal with the New Zealand met service, Rhetorical is supplying rVoice to the country’s MetPhone dial-up forecast service.

Many of New Zealand’s farmers and fishermen rely on MetPhone for accurate forecasts of the country’s "fairly extreme" weather, according to Stephen Harris, operations manager at the met service.

He says: "The rVoice package was the best choice for us because it had the most natural sounding voice, the price was right and our surveys showed people liked it."

Rhetorical has also signed a contract with the Australian group Adacel to supply the rVoice package to help train air traffic controllers in the US Air Force.

Orange is using the firm’s text-to-speech technology to give customers news and regional weather reports. The deal is an upgrade to Orange’s original service, which previously used pre-recorded information, and will use two voices from Rhetorical’s rVoice technology to read the requested information to the user - an English female voice for weather and an American female for news.

Emma Jones, voice portal development manager at Orange, says the group first looked into text-to-speech systems two years ago, but found the quality was not up to scratch.

She adds: "However, on being introduced to Rhetorical’s text-to-speech system we were impressed by the ability of rVoice to read back complicated names and sentences."

Mr Moens says: "This year has been marked by our success in winning contracts over some of the larger text-to-speech providers.

"With an expanding customer base, including some of the largest telecoms firms in the US, we have now secured our position as a key player in the international speech technology market."

The end users of Rhetorical’s products are not required to download any specific software to listen to the voices, which will run on any computer providing it has a sound card, because all the computing work is done at the other end of the line.

Mr Moens says: "One of our clients is Sparda Bank in Germany, which allows customers to interact with the system because it can recognise speech and will ask questions."

"Using speech recognition and text-to-speech means companies can automate a lot of call centre operations, which means customers won’t be on hold for so long and operators can get on with more interesting work."

Rhetorical took on 15 new staff last year, boosting its workforce to 45, and the bulk of those are based in the Capital.

When asked whether the company, which also has offices in London, Germany and Boston, plans to open any more overseas offices, Mr Moens says: "That’s certainly something that could happen."

He adds: "We have just rolled out US Spanish, which is extremely useful for the North American market, where we are already represented by our Boston office, but this could open some doors into the South American market."

"Our overseas expansion could happen by opening new offices or it could be by further developing our partnerships around the world."

While gross profits soared to 1.1m last year, Rhetorical’s net losses were held steady at 1.6m, despite the growing workforce, and the firm is confident of breaking even in around six months.

Tony Robison, the company’s chief financial officer, said: "We have again demonstrated that we can grow revenues and grow the business without an increase in net losses. We believe that we can continue this level of revenue growth with a much lower increase in our cost base."

Rhetorical raised 4.3m in new funding in November 2001, and Mr Moens insists it will not need any additional external cash in the near future.

"We don’t need to go out for any more money before we get to breakeven," he says.

Although a fresh injection of equity has not been formally discussed, Mr Moens says returning to the venture capital market with a profitable company was "certainly a possibility".

Along with many different English voices, such as Scottish, American and Australian, Rhetorical has developed Greek and Spanish language packages.

The software is so sophisticated it can detect whether a word such as "central" should be pronounced in English or Spanish, Mr Moens says.

"Once we’ve done a language like English, setting up new voices is very efficient. New languages are more difficult, but there are general methodologies behind the process.

"All our voices start from the recording of a human voice, preferably at our studio here in Edinburgh. It takes several hours, and then we break that recording down into the building blocks of the language.

"Once you’ve got the voice and give it some text, it can start reading it back immediately, and the computing work is done at 40 times the speed of real time."

One such example is the use of rVoice to generate automated outgoing telephone calls from businesses such as call centres or delivery firms.

And Birmingham-based Truancy Call, which serves around 150 schools across the UK, is using the software to call the parents of children who are not in school. If parents are not aware of their child’s absence, they can push a button that automatically connects them with the school.

Mr Moens says the development of every new language throws up unexpected difficulties "because each one is so different", but the principal of breaking each one down to its component sound structures remains the same.

Having established international recognition for its software, Mr Moens insists there is still much to be done at Rhetorical, which spun out from Edinburgh University three years ago.

"There will be more languages, which will open new markets, but we’ll also work on perfecting our existing voices," he says.

"So far we’ve mainly targeted the telephony market, such as call centres and voice portals, but we see a big market in computer games and other forms of entertainment."

Talented trio who made their voice heard

RHETORICAL was founded by Marc Moens, Paul Taylor and Peter Denyer and was spun out of Edinburgh University three years ago.

Mr Moens has a background in linguistics and computer science, while Mr Taylor, who is the firm’s chief technical officer, has a detailed knowledge of electronics and speech systems.

The company is chaired by Mr Denyer, whom Mr Moens credits with giving the team the confidence to set up in business together.

Mr Denyer is also the chairman of MicroEmissive Displays, which makes tiny electronic screens, and until February was the chairman of the Scottish Microelectronics Centre in Edinburgh.

The founding trio own around 51 per cent of the business, with the rest held by a number of venture capital firms, including Pentech, Scottish Equity Partners, Friends Ivory & Sime, Artemis, Pennine VCT and Capital for Companies.

Among other achievements, Rhetorical provided the voice for Seonaid, the "cyberbabe" who guides youngsters around the Scottish Executive’s official website.