Yinyin Liu
Yinyin Liu has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Harini Eavani
Intel AI
2 shared publications - William Yang Wang
University of California Santa Barbara
2 shared publications - Ana R Grant
Synthetic Genomics Inc
1 shared publication - Anna Bethke
Intel AI Santa Clara CA USA
1 shared publication - Ariel S Schwartz
Synthetic Genomics Inc
1 shared publication - Arjun K. Bansal
Intel AI Lab Santa Clara CA
1 shared publication
Latest Publications
- Open Access
Few-Shot NLG with Pre-Trained Language Model
Neural-based end-to-end approaches to natural language generation (NLG) from structured data or knowledge are data-hungry, making their adoption for real-world applications difficult with limited data. In this work, we propose the new task of \textit{few-shot natural language generation}. Motivated by how humans tend to summarize tabular data, we propose a simple yet effective approach and show that it not only demonstrates strong performance but also provides good generalization across domains. The design of the model architecture is based on two aspects: content selection from input data and language modeling to compose coherent sentences, which can be acquired from prior knowledge. With just 200 training examples, across multiple domains, we show that our approach achieves very reasonable performances and outperforms the strongest baseline by an average of over 8.0 BLEU points improvement. Our code and data can be found at \url{https://github.com/czyssrs/Few-Shot-NLG}
Open AccessFew-Shot NLG with Pre-Trained Language Model
Neural-based end-to-end approaches to natural language generation (NLG) from structured data or knowledge are data-hungry, making their adoption for real-world applications difficult with limited data. In this work, we propose the new task of \textit{few-shot natural language generation}. Motivated by how humans tend to summarize tabular data, we propose a simple yet effective approach and show that it not only demonstrates strong performance but also provides good generalization across domains. The design of the model architecture is based on two aspects: content selection from input data and language modeling to compose coherent sentences, which can be acquired from prior knowledge. With just 200 training examples, across multiple domains, we show that our approach achieves very reasonable performances and outperforms the strongest baseline by an average of over 8.0 BLEU points improvement. Our code and data can be found at \url{https://github.com/czyssrs/Few-Shot-NLG}
- Open Access
- 1 January 2019
- Published by Association for Computational Linguistics (ACL)
- p. 4757-4766
A Benchmark Dataset for Learning to Intervene in Online Hate Speech
Jing Qian, Anna Bethke, Yinyin Liu, Elizabeth Belding, William Yang Wang. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.
- 1 January 2019
- Published by Association for Computational Linguistics (ACL)
- p. 4757-4766
Open AccessA Benchmark Dataset for Learning to Intervene in Online Hate Speech
Jing Qian, Anna Bethke, Yinyin Liu, Elizabeth Belding, William Yang Wang. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.
- Open Access
Deep Semantic Protein Representation for Annotation, Discovery, and Engineering
Computational assignment of function to proteins with no known homologs is still an unsolved problem. We have created a novel, function-based approach to protein annotation and discovery called D-SPACE (Deep Semantic Protein Annotation Classification and Exploration), comprised of a multi-task, multi-label deep neural network trained on over 70 million proteins. Distinct from homology and motif-based methods, D-SPACE encodes proteins in high-dimensional representations (embeddings), allowing the accurate assignment of over 180,000 labels for 13 distinct tasks. The embedding representation enables fast searches for functionally related proteins, including homologs undetectable by traditional approaches. D-SPACE annotates all 109 million proteins in UniProt in under 35 hours on a single computer and searches the entirety of these in seconds. D-SPACE further quantifies the relative functional effect of mutations, facilitating rapid in silico mutagenesis for protein engineering applications. D-SPACE incorporates protein annotation, search, and other exploratory efforts into a single cohesive model.
Open AccessDeep Semantic Protein Representation for Annotation, Discovery, and Engineering
Computational assignment of function to proteins with no known homologs is still an unsolved problem. We have created a novel, function-based approach to protein annotation and discovery called D-SPACE (Deep Semantic Protein Annotation Classification and Exploration), comprised of a multi-task, multi-label deep neural network trained on over 70 million proteins. Distinct from homology and motif-based methods, D-SPACE encodes proteins in high-dimensional representations (embeddings), allowing the accurate assignment of over 180,000 labels for 13 distinct tasks. The embedding representation enables fast searches for functionally related proteins, including homologs undetectable by traditional approaches. D-SPACE annotates all 109 million proteins in UniProt in under 35 hours on a single computer and searches the entirety of these in seconds. D-SPACE further quantifies the relative functional effect of mutations, facilitating rapid in silico mutagenesis for protein engineering applications. D-SPACE incorporates protein annotation, search, and other exploratory efforts into a single cohesive model.