A Linguistic Approach to Short Query and Answer Systems

Conversational agents for natural language interaction, in general, face some challenges in providing content to support the responses requested by users. Usually of the content of the responses is manually defined, which is not an adequate solution. This work presents an approach to enabling these systems to use automatically the concepts and relationships stored in Knowledge Bases. The main contribution of our approach is the generalization of the system, which can answer questions related to multiple domains. This is accomplished based on linguistic information as support to the activities of natural language understanding. The results obtained in the classification of question types and identification of nominal complements are presented and indicate promising results.