Towards alignment strategies in human-agent interactions based on measures of lexical repetitions

Alignment of communicative behaviour is an important feature Human–Human interaction that directly affects the collaboration and social connection conversational partners. With aim improving abilities a virtual agent, in particular its strategies related to (lexical) verbal alignment, this article focuses on alignment linguistic productions dialogue participants task-oriented dialogues. We propose new framework quantify both lexical self-repetition behaviours from dyadic transcripts. The involves easily computable measures based repetition patterns automatically extracted via sequential pattern mining approach. These allow characterisation nature these processes by addressing various informative aspects such as their variety, complexity, strength. This implemented freely available open-source software dialign. Using measures, we present contrastive study between Human–Agent dialogues corpora reveals major differences behaviours. Lastly, address challenge integrating capabilities artificial agents. To end, describe guidelines discuss integration proposed real-time system.