BlueSkyMARCUS: An Event-Centric NLP Pipeline that generates Character Arcs from Narratives
Published in Text2Story Workshop @ ECIR 2022, 2022
Recommended citation: Bhyravajjula, S., Narayan, U., & Shrivastava, M. (2022). "MARCUS: An Event-Centric NLP Pipeline that generates Character Arcs from Narratives." Text2Story @ ECIR 2022. https://ceur-ws.org/Vol-3117/paper7.pdf
| 🏆 Best Paper Award | Student Travel Grant by IIIT-H |
Built a novel pipeline that (1) used a BiLSTM tagger to label events and a Semantic Role Labeller to identify participant characters from long-form narratives, (2) used fine-tuned BERT and RoBERTa models to extract proxy markers of emotion and sentiment to measure a character’s circumstance, and (3) plotted smoothed character arcs as an aggregated change of character circumstance across event sequences.
Authors: Sriharsh Bhyravajjula, Ujwal Narayan, Manish Shrivastava
Venue: Text2Story @ ECIR 2022, Stavanger, Norway
Recommended citation: Bhyravajjula, S., Narayan, U., & Shrivastava, M. (2022). “MARCUS: An Event-Centric NLP Pipeline that generates Character Arcs from Narratives.” Text2Story @ ECIR 2022.