The Interpretive-Schematic conceptual framework: Grounded interpretation in NLP through an ethnographic lens

The implied universalist and formalist semantic assumptions in today's Natural Language Processing (NLP) paradigms used in artificial intelligence systems fundamentally lack the capacity to accommodate subjective grounded interpretations of written language; assumptions that offer seemingly practical results, while encouraging a biased and objective view of reality. This thesis attempts to identify such underlying theoretical assumptions with respect to meaning, and investigates the adequacy of the semantic representations developed based on them. In order to suggest an alternative approach, through a hermeneutic phenomenological review of a selected literary corpus, the possibility of translating interpretive ethnography into NLP has been explored. In the end, the capacity of image schemata as a cognitive linguistic tool in facilitating this translation is studied and a conceptual framework is developed to demonstrate its descriptive and interpretive potential in a non-universalist approach towards NLP. The inherent subjectivity of image schemata and their computational models in NLP are two major limits of this approach which could be the subject of further research.