The final speaker in this final Social Media & Society 2024 session is my QUT colleague Kate O’Connor Farfan, whose interest is in the use of semiotics in combination with Natural Language Processing (NLP) for the study of polarisation. NLP comes with a very diverse range of applications, variously examining superficial and structural aspects at differing levels of complexity.
Kate’s work is interested centrally in the structure of texts, and dependency parsing is a useful tool for this – but such analytical frameworks also substantially complicate the analysis: dependency parsing can show up some 40 or more relationships between words, for instance. So, the approach here is to operationalise polarisation by identifying the language configurations from semiotics theory that speak to the characteristics of polarisation.
If dependency parsing largely deconstructs sentences into their constituent parts, the challenge here is to determine which of these parts relate to the aspects of polarisation that may be operationalised. Similarly, if Named Entity Recognition identifies the various actors being named, the challenge here is what connotations these names carry.
A core relationship here is what dependency parsing refers to as nsubj: the noun which is the subject of a verb. The systematic study of such relationships, especially for verbs which express the desires of the speaker, can reveal how speakers understand themselves and others, how they relate to specific issues, topics, and actors, and what functions they ascribe to such entities.
Semiotic theory can guide our selection and organisation of such NLP methods in order to trace the characteristics that can reveal the aspects of a given phenomenon.
