Command Palette
Search for a command to run...
Long N. Phan; James T. Anibal; Hieu Tran; Shaurya Chanana; Erol Bahadroglu; Alec Peltekian; Grégoire Altan-Bonnet
Abstract
In this report, we introduce SciFive, a domain-specific T5 model that has been pre-trained on large biomedical corpora. Our model outperforms the current SOTA methods (i.e. BERT, BioBERT, Base T5) on tasks in named entity relation, relation extraction, natural language inference, and question-answering. We show that text-generation methods have significant potential in a broad array of biomedical NLP tasks, particularly those requiring longer, more complex outputs. Our results support the exploration of more difficult text generation tasks and the development of new methods in this area
Code Repositories
Benchmarks
| Benchmark | Methodology | Metrics |
|---|---|---|
| document-classification-on-hoc | SciFive-large | F1: 86.08 |
| drug-drug-interaction-extraction-on-ddi | SciFive-large | F1: 0.8367 Micro F1: 83.67 |
| named-entity-recognition-ner-on-jnlpba | SciFive-Large | F1: 77.55 |
| named-entity-recognition-ner-on-ncbi-disease | SciFive-Base | F1: 89.39 |
| named-entity-recognition-on-bc5cdr-chemical | SciFive-Large | F1: 94.76 |
| named-entity-recognition-on-bc5cdr-disease | SciFive-Large | F1: 87.62 |
| named-entity-recognition-on-species-800 | SciFive-Base | F1: 76.55 |
| natural-language-inference-on-mednli | SciFive-large | Accuracy: 86.57 Params (M): 738 |
| relation-extraction-on-chemprot | SciFive Large | F1: 78 |
| relation-extraction-on-chemprot | BioT5X (base) | F1: 77.40 |
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.