EMNLP 2016 (Conference on Empirical Methods in Natural Language Processing), a top conference in natural language processing, was held on November 1-5, 2016 in Austin, Texas, USA. Three papers from Prof. Kewei Tu’s research group at SIST were accepted by the conference. Prof. Kewei Tu, along with the two first-year PhD students from the group, Yong Jiang and Wenjuan Han were invited to attend the conference and presented the papers.
A brief description of the three papers:
- Lin Qiu, Kewei Tu and Yong Yu, "Context-Dependent Sense Embedding". In this paper, a novel probabilistic model for embedding word senses in a continuous space using context information is proposed. The model outperforms the state-of-the-art model on a word sense induction task by a 13% relative gain.
- Yong Jiang, Wenjuan Han and Kewei Tu, "Unsupervised Neural Dependency Parsing". This paper proposes to use a neural model to predict grammar rule probabilities based on distributed representation of POS tags. The model outperforms previous approaches in experiments with nine different languages.
- Kewei Tu, "Modified Dirichlet Distribution: Allowing Negative Parameters to Induce Stronger Sparsity". This paper proposes a simple modification to the Dirichlet distribution, which induces stronger sparsity as a prior.
EMNLP is one of the three top conferences (ACL, EMNLP, and NAACL) in natural language processing. The EMNLP 2016’s valid submissions was 1087 and the acceptance rate was 24.3%.