Continuous Speech Recognition at Hellas Flying Dolphins Call Centre
Hellas Flying Dolphins (HFD), one of the largest and most successful maritime companies in Greece, aims at improving coastal transportation for conventional ships and new technology high-speed vessels. The company employs 3, 000 crew and other staff, and does business with over 1.100 Greek suppliers, which places it among top companies not only locally, but also in Europe.
However, the high number of calls fielded by the company's call centre - especially during summer time - emphasised the need for a high quality technology solution to help maintain customer service levels, and to improve call centre operations. Investment in new technologies is of great importance to HFD, which sought to fully exploit voice recognition and other technologies designed to improve call centre throughput and service.
The company aimed to develop a dynamic partly automated call centre - available all day, every day - that offered a range of maritime information services on destinations, itineraries, and ticket availability. By implementing voice technologies, HFD aimed to operate a call centre that would be capable of scaling to large call volumes, and of always delivering quick and direct customer service even during peak periods.
After a market survey of available options, tools and solutions, HFD chose to use the services of Knowledge S.A., a Greek company with long experience in language technology, and particularly in voice technologies.
Expert Speech Technology Providers
Such systems "listen to" users, "understand" their requests and respond by taking appropriate actions. Available solutions are based on the company's "Oral Dialogue Systems" platform and on SpeechMania, a Philips Speech Processing system. The combined technologies of the two companies allows for systems which recognise an almost infinite number of words, without requiring certain key words or specific dialogue models. Users may interact in the same way they would do with a person, using natural language, regional accents, incomplete sentences, or even making grammatical mistakes.
Philips SpeechMania platform and Intel's CT Connect were selected for the HFD call centre. The Knowledge team worked on the development of the appropriate dialogue handling, customised the system, and integrated all of the tools. During the development process, no specific problems were encountered and co-operation between the two companies was remarkably smooth. The fact that both companies were partners in the EU-funded E2M project may well have been a contributory factor.
In October 2002, the call centre was made available to the general public in a test period that lasted until the end of the year. Since January 2003, anyone who needs information on HFD itineraries and dials + may use the automatic call centre, which is now in full use. The company has also implemented a special prompt to enable users to request the transfer of their call to a human agent if they so prefer.
According to call statistics, around 26% of incoming calls were directly connected to an agent, perhaps indicating that those users still do not have much confidence in the technology. However, around 55% of the total calls are successfully serviced automatically, 91% of which are fully managed by the system. Missed calls due to system malfunctions make up less than 3,5% of the total number, which is a significant gain.
Conclusions & Future Plans
At the same time, the quality of service is of a very high standard and recognition rates reach 95%, which contributes to natural interaction and ease-of-use factors. Antonis Fotinos notes: "The system had not been online for long when a customer, after interaction with the automatic system, was connected to an agent and said: "Well, the girl I was talking to a while ago, said that there are free tickets to Mykonos. I'd like to make a reservation!." He had not realised that he was actually talking to a machine!"
A second phase of the project to expand the call centre to support full booking services is under consideration. When the system recognises a yes-answer in the question "Do you want to make a reservation?", a unique reservation number will be stored in the database. Using this number, customers will be able to buy tickets from any of the groups tourist agencies. The automation of the this procedure, from simple information provision until the purchase of a ticket, is estimated to be a further boost for the company.
According to Antonis Fotinos, "Most companies wish to follow the example of EasyJet, that is, to automate the booking procedure to the maximum possible degree, at the same time minimising the company's operational cost." It is a pattern that has been successful for many businesses around the world and it was made possible basically due to the availability the necessary technological platforms and tools.
This success story is based on an interview given by Mr Antonis Fotinos from the Language Engineering Dept/Pre-sales of Knowledge S.A on February 25th 2003. We would like to thank him very much for his co-operation and assistance.