An ensemble framework for sentiment-embedded event evolution in diaspora oral archives

Introduction. Diaspora oral archives should be displayed in the context of diversity and inclusiveness rather than being glossed over by the dominant, normative group so their voices need to be spread further. Method. This paper proposes an ensemble framework for sentiment-embedded event evolution in diaspora oral archives. It contains a knowledge representation model, an event evolutionary graph, and an event extraction workflow to extract entities, events, and relationships. Results. The South Asian Oral History Project is selected as the data source. The key events, entities, event types, sentiment and event relations are extracted with natural language processing techniques to construct sentiment-embedded event evolutionary graph. Based on this, the event evolution, spatio-temporal and spatio-sentiment patterns are analysed. Conclusions. Such methods allow researchers and archivists to engage in research on machine-assisted oral archives to ensure reproducibility, reduce interpretative biases, and efficiently and swiftly amplify hidden voices of ‘the other’.