Abstract

In the past year the GAIA team has improved our end-to-end knowledge extraction, grounding, inference, clustering and hypothesis generation system that covers all languages (English, Russian and Ukrainian), data modalities and knowledge element types defined in new AIDA ontologies. We participated in the evaluations of all tasks within TA1, TA2 and TA3 and achieved highly competitive performance. Our TA1 system achieves top performance at both intrinsic evaluation and extrinsic evaluation through TA2 and TA3. The system incorporates a number of impactful and fresh research innovations:
• Attentive Fine-Grained Entity Typing Model with Latent Type Representation: We propose a fine-grained entity typing model with a novel attention mechanism and a hybrid type classifier. We advance existing methods in two aspects: feature extraction and type prediction. To capture richer contextual information, we adopt contextualized word representations instead of fixed word embeddings used in previous work. In addition, we propose a two-step mentionaware attention mechanism to enable the model to focus on important words in mentions and contexts. We also develop a hybrid classification method beyond binary relevance to exploit type interdependency with

Date

2019

Authors

Manling Li, Ying Lin, Ananya Subburathinam, Spencer Whitehead, Xiaoman Pan, Di Lu, Qingyun Wang, Tongtao Zhang, Lifu Huang, Heng Ji, Alireza Zareian, Hassan Akbari, Brian Chen, Bo Wu, Emily Allaway, Shih-Fu Chang, Kathleen R McKeown, Yixiang Yao, Jennifer Chen, Eric Berquist, Kexuan Sun, Xujun Peng, Ryan Gabbard, Marjorie Freedman, Pedro A Szekely, TK Satish Kumar, Arka Sadhu, Ram Nevatia, Miguel E Rodríguez, Yifan Wang, Yang Bai, Ali Sadeghian, Daisy Zhe Wang

Conference

TAC