4Socio-emotional Interaction Strategies: the Case of Alignment
The opinion analysis systems presented in Chapter 3 were developed in the framework of human interactions with an embodied conversation agent. Their purpose is to provide material for agents’ socio-emotional interaction strategies, first by triggering strategies (for example, if the user expresses a given opinion, the agent generates an utterance in response to this opinion), and second as a means of measuring the impact of an interaction strategy on the user (for example, the effects of a given strategy on user opinions).
In this chapter, we shall consider the first of these aspects, developing a strategy to select and generate user utterances in response to a user opinion.
We shall focus on the first of the research questions defined in the introduction, section I.1.3, concerning the relevance of different theoretical models in constructing a computational (linguistic) model by investigating the concept of alignment in terms of appreciation. The strategy chosen here concerns alignment in terms of verbal content. In this context, we shall consider the role of alignment strategies in user engagement (research question Q3, defined in I.3.3).
The computational models presented in this chapter were developed from conversational analysis and psycholinguistic literature and through the analysis of different corpora. Both qualitative and quantitative analysis was carried out for the alignment model presented here.
Get Opinion Analysis in Interactions now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.