Exploring the Drawbacks of Using Artificial Intelligence in Studies on Architecture:NLP and Highly Contextual Postmodern Architectural Discourse

Abstract

This research investigates the artificial intelligence aided architectural discursive sentiment analysis within the texts of the representants of architectural culture of the sixties and the seventies. The subject of inquiry for this study is to examine the nature and validity of sentiment analysis, using Natural Language Processing tools, on postmodernist architectural discourse. This paper searches to identify the possible drawbacks of using the NLP tool. The results show that the sentiment analysis of postmodernist texts using opinion mining techniques must be interpreted with the utmost caution. Especially in the case of unstraightforward, highly contextualized, and modernist reactive, often ironic or sarcastic, postmodern discourse. Two main reasons for that were found in this study. First, architectural postmodernists use many polarized words which might end in inconclusive results when opinion mining techniques are in use. Second, the postmodernist texts might themselves lack important contextual cues for appropriate classification of sentiment by NLP. Discrepancies and internal contradictions are characteristic for the postmodernist style of writing. and artificial intelligence still does not fully decode it.