Papers by Matthew Sims
Measuring Information Propagation in Literary Social Networks (2020.emnlp-main)
Copied to clipboard
| Challenge: | a gap in computational work to support the "Miss Havisham is dead" "She died" research focuses on the representation of social networks in literature . |
| Approach: | They propose a pipeline for measuring information propagation in literature . they analyze the dynamics of information propagations in over 5,000 works of fiction . |
| Outcome: | The proposed pipeline analyzes the dynamics of information propagation in 5,000 works of fiction and finds that women fill structural holes connecting different communities more frequently than men. |
Attending to Long-Distance Document Context for Sequence Labeling (2020.findings-emnlp)
Copied to clipboard
| Challenge: | UC Berkeley researchers develop a method for incorporating global context in long documents . many of the main datasets used in NLP are comprised of relatively short documents - english OntoNotes contains 223 tokens . |
| Approach: | They propose a method for incorporating global context in long documents . they use multiple mentions of the same word type to generate a representation for each token . |
| Outcome: | The proposed model performs better at recognizing entities with high TF-IDF scores than parametric models lacking context. |
Literary Event Detection (P19-1)
Copied to clipboard
| Challenge: | a new dataset of literary events is presented to examine the nature of narratives . literature presents a number of challenges for existing systems, including complex narration . |
| Approach: | They propose a dataset of literary events that are depicted as taking place within the imagined space of a novel. |
| Outcome: | The proposed model achieves an F1 score of 73.9 for prestige and popularity . the best performing model achieve a score of 79.9 for prestige compared to the previous model . |