| dc.description.abstract | The purpose of this thesis was to create a conversational AI for serving fact-checks, using a collection of already existing fact-checking articles. The conversational AI uses a hybrid system, combining both a question answering agent, chitchat agent, and multiple non-AI based skills to perform the task.
The program created consists of a user interface, broker, and seven different skills. For the implementation multiple existing pre-trained deep learning models were used, where many are based on the Transformer architecture. Already fine-tuned versions of these models were used. The conversational AI can present fact-checking articles in multiple ways, fact-check a claim presented, and has some multi-turn capabilities.
The result is a functional conversational AI which is capable of serving fact-checks from a collection of fact-checking articles. Although the conversational AI is functional, there are several issues that should be addressed, and further work to be done. | |
