Alex: An Artificial Conversational Agent for Students at the TU Berlin


User interfaces of most data-centered software systems are oriented on the data structure of the respective application. While such interfaces work well for trained users, they are most often not suitable for non-expert users. In this paper, we present ALEX, an advisory artificial conversational agent that uses machine learning and natural language processing to gather information from a relational database. Specifically, ALEX allows students at the Technische Universität Berlin (TU Berlin) to query courses and modules in a conversational way using textual input. The natural language understanding of the system is realized with a pipeline that uses a Hidden Markov Model tagger to annotate the input and groups phrases together to form SQL queries. ALEX was evaluated in a research study that compared the conversational agent with the existing online systems of the TU Berlin which are currently used by students to retrieve information about courses. The results show that the conversational approach is not only more efficient with nonexpert users, but also has higher hedonic and pragmatic quality. A live demo of the tool can be found at

Year: 2017
In session: Poster
Pages: 238 to 245