Li et al. (2016)[1]: infer sense embeddings from sense definition. Steps: 1/ train RNN on monosemous words, assume its sense embedding is the same as its word embeddings, 2/ apply to polysemous words.
References[]
- ↑ Li, Qi, Tianshi Li, and Baobao Chang. "Learning Word Sense Embeddings from Word Sense Definitions." arXiv preprint arXiv:1606.04835 (2016).