ConclusionSummary and Directions for Future Research

The research presented in this book touches on a variety of domains in an attempt to define methodologies for opinion analysis applications.

In the context of affective computing, our work relates specifically to the way in which verbal content is taken into account in analyzing opinion-linked phenomena and more broadly, for socio-emotional behaviors – traditionally approached in terms of acoustic and facial cues. We compared theoretical models developed by the natural language processing community with those developed in affective computing; this was a particularly complex task, given the areas of overlap and the subtle differences between the theories in question.

In the domain of natural language processing, our contribution relates to the definition of opinion-linked linguistic phenomena and their characterization in in-the-wild data, both in human–human or human–machine interactions and written or oral contexts.

We presented work on the characterization of spontaneous speech phenomena, and on the enunciative constructs specific to verbal communications and the particular way they are formed. We carried out a detailed study of these phenomena based on different theoretical models from the fields of psychology, linguistics and marketing.

In order to detect these phenomena, we worked on a wide variety of rule-based and machine learning-based methods, in the hope of creating a hybrid method combining the best elements of ...

Get Opinion Analysis in Interactions now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.