Character-level Representations Improve DRS-based Semantic Parsing Even in the Age of BERT
| Autor: | Antonio Toral, Rik van Noord, Johan Bos |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
FOS: Computer and information sciences
Semantic HTML Computer science 02 engineering and technology Representation (arts) Semantic parsing computer.software_genre Discourse Representation Structures 03 medical and health sciences 0302 clinical medicine Rule-based machine translation 0202 electrical engineering electronic engineering information engineering Structure (mathematical logic) Parsing Computer Science - Computation and Language Character-level models business.industry Character (mathematics) 030221 ophthalmology & optometry 020201 artificial intelligence & image processing Language model Artificial intelligence business Encoder computer Computation and Language (cs.CL) Natural language processing |
| Zdroj: | Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 4587-4603 STARTPAGE=4587;ENDPAGE=4603;TITLE=Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) EMNLP (1) |
| ISSN: | 4587-4603 |
| Popis: | We combine character-level and contextual language model representations to improve performance on Discourse Representation Structure parsing. Character representations can easily be added in a sequence-to-sequence model in either one encoder or as a fully separate encoder, with improvements that are robust to different language models, languages and data sets. For English, these improvements are larger than adding individual sources of linguistic information or adding non-contextual embeddings. A new method of analysis based on semantic tags demonstrates that the character-level representations improve performance across a subset of selected semantic phenomena. EMNLP 2020 (long) |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|