Lample, G., Ballesteros, M., Subramanian, S., Kawakami, K., & Dyer, C. (2016). Neural Architectures for Named Entity Recognition. https://doi.org/10.18653/v1/N16-1030
Research Gap
State-of-the-art named entity recognition systems rely heavily on hand-crafted features and domain-specific knowledge in order to learn effectively from the small, supervised training corpora that are available.
Research Work
They introduce two new neural architectures
- one based on bidirectional LSTMs and conditional random fields
- one that constructs and labels segments using a transition-based approach inspired by shift-reduce parsers.
Their model rely on two sources of information about words:
- character-based word representations learned from the supervised corpus.
- unsupervised word representations learned from unannotated corpora.
Token-level evidence for "being a name" includes both
-
orthographic evidence (what does the word being tagged as a name look like?)
- use character-based word representation model (Ling et al., 2015b) to capture orthographic sensitivity.
-
distributional evidence (where does the word being tagged tend to occur in a corpus)
- use distributional representations (Mikolov et al., 2013b) to capture distributional sensitivity.
Results
obtain state-of-the-art performance in NER in four languages
Innovation
without resorting to any language-specific knowledge or resources such as gazetteers.
