A General Framework for Information Extraction using Dynamic Span Graphs

We introduce a general framework for several information extraction tasks that share span representations using dynamically constructed span graphs. The graphs are constructed by selecting the most confident entity spans and linking these nodes with confidence-weighted relation types and coreferences. The dynamic span graph allows coreference and relation type confidences to propagate through the graph to iteratively refine the span representations. This is unlike previous multi-task frameworks for information extraction in which the only interaction between tasks is in the shared first-layer LSTM. Our framework significantly outperforms the state-of-the-art on multiple information extraction tasks across multiple datasets reflecting different domains. We further observe that the span enumeration approach is good at detecting nested span entities, with significant F1 score improvement on the ACE dataset.

PDF Abstract NAACL 2019 PDF NAACL 2019 Abstract

Results from the Paper

 Ranked #1 on Relation Extraction on ACE 2004 (Cross Sentence metric)

     Get a GitHub badge
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Relation Extraction ACE 2004 DyGIE RE Micro F1 59.7 # 4
NER Micro F1 87.4 # 5
Cross Sentence Yes # 1
Relation Extraction ACE 2005 DyGIE RE Micro F1 63.2 # 8
NER Micro F1 88.4 # 8
Sentence Encoder ELMo # 1
Cross Sentence Yes # 1
Joint Entity and Relation Extraction SciERC DyGIE Entity F1 65.2 # 9
Relation F1 41.6 # 6
Cross Sentence Yes # 1
Named Entity Recognition WLPC DyGIE F1 79.5 # 1
Relation Extraction WLPC DyGIE F1 64.1 # 2