Papers by Shou-de Lin
Glyph2Vec: Learning Chinese Out-of-Vocabulary Word Embedding from Glyphs (2020.acl-main)
Copied to clipboard
| Challenge: | Chinese NLP applications that rely on large text often contain huge amounts of vocabulary which are sparse in corpus. |
| Approach: | They propose a multi-modal model that extracts visual features from Chinese word glyphs to expand current word embedding space without accessing any corpus. |
| Outcome: | The proposed model can embed words in Chinese without accessing corpus without a corpus. |
Explaining Word Embeddings via Disentangled Representation (2020.aacl-main)
Copied to clipboard
| Challenge: | Disentangled representations are known to represent interpretable factors in separated dimensions. |
| Approach: | They propose to transform dense word vectors into disentangled embeddings with improved interpretability by encoding polysemous semantics separately. |
| Outcome: | The proposed model can be encoded into multiple sub-embeddings or sub-areas and generates more efficient and effective features for natural language processing. |
Explainable and Sparse Representations of Academic Articles for Knowledge Exploration (2020.coling-main)
Copied to clipboard
| Challenge: | a system for summarizing academic articles by concept tagging has shown great coverage and high accuracy of concept identification. |
| Approach: | They propose to transform tagged concepts into sparse vectors as representations of academic documents. |
| Outcome: | The proposed system can be applied to a broader class of applications. |