Same text, same discourse? Empirical validation of a discourse analysis methodology for cultural heritage

Abstract

Abstract Most of the new knowledge about our past that is generated in cultural heritage disciplines is produced in non-structured forms, such as project reports, monographs, or research papers. This proliferation of cultural heritage knowledge in textual discourse form requires support for information structuration and extraction of the semantic relations embedded in these texts, in order to fully understand and post-process them. Previous works show that discourse analysis techniques can represent pieces of textual information in a highly structured manner, identifying the semantic relations between them and the entities referred to in the discourse according to well-established linguistic patterns. This is particularly useful for aligning the structure and information extracted with information system post-processing techniques, such as natural language processing, that are currently applied to cultural heritage. In this paper, we apply discourse models through a previously developed methodology to a real case study to represent and analyse cultural heritage textual narratives. Then, we perform an empirical study with cultural heritage experts in order to validate the models obtained, the resultant information, some usability aspects of the approach, and the degree of agreement between experts. The main objective was to obtain empirical and quantitative validation of the application of our proposed discourse analysis methodology to narratives in cultural heritage, its implications for different specialists, and its possibilities as a methodology to assist in providing structure to and capturing the semantics of freestyle documents, to automatize some discourse information extraction tasks and to facilitate their post-processing.