Abstract:

Nowadays, much is the work that remains to be done in the adoption of conversational agents, because a very expensive process is needed to adapt them to a specific domain. In addition, they are still conditioned to the questions asked by the user, often unpredictable. So designing the conversation flows are very important in the development process of these agents. Even so, conversation agents have been implanted in many domains, such as commerce, customer service, health, science, education, etc.

Continuing with the application of agents within the context of education, this proposal presents Test-Bot, a prototype designed to evaluate objective tests in real time. It has been trained with a corpus formed by 425 questions and answers extracted from the different tests carried out in previous years in the subject Information Systems and Technology, a subject taught in the Computing Engineering degree.

From a corpus made up of 425 questions and answers the different conversational flows have been designed, which it has been adapted according to the answers given by users; So, the contents that the student doesn’t have properly secured are reinforced; the training phrases have been extracted and classified according to the type of content of the questions, finally, the different intents have been identified.

The evaluation test questions, contains a lot of information to provide knowledge to the conversational agent. The use of conversational agents in the realization of objective tests, allows the student to perform the tests autonomously, obtaining feedback at the time. The conversational agent is presented as a tool to support the student seeking help in the learning process, and teachers allow them to see the evolution of the student over time.

Keywords:

Chatbot, artificial intelligence, natural language processing, objective tests.