A5045560420- Human Pose and Action Recognition
- Video Analysis and Summarization
- Text Readability and Simplification
- Multimodal Machine Learning Applications
- Sports Analytics and Performance
- Natural Language Processing Techniques
- Topic Modeling
École Polytechnique Fédérale de Lausanne
2019-2022
The paper describes the submission of team "We used bert!" to shared task Gendered Pronoun Resolution (Pair pronouns their correct entities). Our final model based on fine-tuned BERT (Bidirectional Encoder Representations from Transformers) ranks 14th among 838 teams with a multi-class logarithmic loss 0.208. In this work, contribution transfer learning technique pronoun resolution systems is investigated and gender bias contained in classification models evaluated.
This paper investigates the modeling of automated machine description on sports video, which has seen much progress recently. Nevertheless, state-of-the-art approaches fall quite short capturing how human experts analyze scenes. There are several major reasons: (1) The used dataset is collected from non-official providers, naturally creates a gap between models trained those datasets and real-world applications; (2) previously proposed methods require extensive annotation efforts (i.e.,...