Multi-lingual dialogue systems for eCommerce
The days when customers had to queue to enjoy the services of their banks or travel agencies have long gone. An advanced speech recognition dialogue system can now identify the customer's voice and allows them to carry out bank transactions safely, order payments or even confirm their travel times over the phone in the language of their choice. Demonstrators of the project have developed sensitised, personalised dialogue systems that use personal data such as the customers' history, behaviour, age, gender and income. Messaging for instance, features a type of voice and language compatible with the user's personality, i.e. direct, open, simple or formal. Frequent use of the system by the user can enrich the user's profile and enhances the system's identification capacity, as it records the user's accent, use of grammar, style of speech and language. Dialogue demonstrators also allow users to customise their banking services to their own needs. They can order foreign currency and choose from having it delivered to their home address to collecting it from their local branch. They can handle train timetable information both for long-distance and regional train routes in Europe and allow the user to give any or all of the journey details from the departure and destination stations to travel dates and timetables all in a single input utterance. In addition, they can recognise and confirm spoken credit card details like card number and an expiry dates for automated payments of a particular order. Dialogue systems are well flexible and adaptive. They are able to file possible responses into various adaptation triggers and observe their range and degree. In cases such as a call altering or a turn-taking conversation pause, the system recovers the error and the dialogue continues uninterrupted. Currently, large German, Italian and the British enterprises successfully make use of the system in their e-customer services in their own language. The ultimate goal of the project is to create advanced spoken HLT user interfaces for mass-market consumer applications also via the Internet.
