Content Finder AssistanT

This paper focuses on ” how to design and develop a dialogue system with a minimal effort “. It presents a novel model for automatic generation of dialogue systems built from contents. This approach is similar to and relies on a search engine, but with augmented dialogue capabilities: at each dialogue turn, the system propose n keywords, in order to optimise the information gain expectation. Its implementation, CFAsT, endeavours to keep the best from both worlds: the universality and automatic generation from search engines, and the usability, the assistance and the self optimisation provided by the dialogue systems. Thus, a beta dialogue application can be generated with no effort. It may serve to gather a dialogue corpus, or to gain a first return of experience. It can also be used as a low cost service. Afterwards, it can be improved with dedicated dialogue strategies.