Matches in SemOpenAlex for { <https://semopenalex.org/work/W3092696934> ?p ?o ?g. }
Showing items 1 to 97 of
97
with 100 items per page.
- W3092696934 abstract "Transformer-based pre-trained language models (PLMs) have dramatically improved the state of the art in NLP across many tasks. This has led to substantial interest in analyzing the syntactic knowledge PLMs learn. Previous approaches to this question have been limited, mostly using test suites or probes. Here, we propose a novel fully unsupervised parsing approach that extracts constituency trees from PLM attention heads. We rank transformer attention heads based on their inherent properties, and create an ensemble of high-ranking heads to produce the final tree. Our method is adaptable to low-resource languages, as it does not rely on development sets, which can be expensive to annotate. Our experiments show that the proposed method often outperform existing approaches if there is no development set present. Our unsupervised parser can also be used as a tool to analyze the grammars PLMs learn implicitly. For this, we use the parse trees induced by our method to train a neural PCFG and compare it to a grammar derived from a human-annotated treebank." @default.
- W3092696934 created "2020-10-22" @default.
- W3092696934 creator A5009993083 @default.
- W3092696934 creator A5018370260 @default.
- W3092696934 creator A5043883660 @default.
- W3092696934 creator A5054936589 @default.
- W3092696934 date "2020-10-19" @default.
- W3092696934 modified "2023-09-27" @default.
- W3092696934 title "Heads-up! Unsupervised Constituency Parsing via Self-Attention Heads" @default.
- W3092696934 cites W1495446613 @default.
- W3092696934 cites W1502293651 @default.
- W3092696934 cites W1970961429 @default.
- W3092696934 cites W2047706513 @default.
- W3092696934 cites W2152907450 @default.
- W3092696934 cites W2888844359 @default.
- W3092696934 cites W2906152891 @default.
- W3092696934 cites W2910243263 @default.
- W3092696934 cites W2912206855 @default.
- W3092696934 cites W2914924671 @default.
- W3092696934 cites W2932637973 @default.
- W3092696934 cites W2948967934 @default.
- W3092696934 cites W2950813464 @default.
- W3092696934 cites W2954366630 @default.
- W3092696934 cites W2964303116 @default.
- W3092696934 cites W2965373594 @default.
- W3092696934 cites W2970862333 @default.
- W3092696934 cites W2971418718 @default.
- W3092696934 cites W3006881356 @default.
- W3092696934 cites W3013891696 @default.
- W3092696934 cites W3023063853 @default.
- W3092696934 cites W3034503989 @default.
- W3092696934 cites W3035305735 @default.
- W3092696934 cites W3035390927 @default.
- W3092696934 cites W3092302419 @default.
- W3092696934 cites W3174985167 @default.
- W3092696934 hasPublicationYear "2020" @default.
- W3092696934 type Work @default.
- W3092696934 sameAs 3092696934 @default.
- W3092696934 citedByCount "0" @default.
- W3092696934 crossrefType "posted-content" @default.
- W3092696934 hasAuthorship W3092696934A5009993083 @default.
- W3092696934 hasAuthorship W3092696934A5018370260 @default.
- W3092696934 hasAuthorship W3092696934A5043883660 @default.
- W3092696934 hasAuthorship W3092696934A5054936589 @default.
- W3092696934 hasConcept C119857082 @default.
- W3092696934 hasConcept C121332964 @default.
- W3092696934 hasConcept C154945302 @default.
- W3092696934 hasConcept C165801399 @default.
- W3092696934 hasConcept C177264268 @default.
- W3092696934 hasConcept C186644900 @default.
- W3092696934 hasConcept C199360897 @default.
- W3092696934 hasConcept C204321447 @default.
- W3092696934 hasConcept C206134035 @default.
- W3092696934 hasConcept C41008148 @default.
- W3092696934 hasConcept C53893814 @default.
- W3092696934 hasConcept C62520636 @default.
- W3092696934 hasConcept C66322947 @default.
- W3092696934 hasConceptScore W3092696934C119857082 @default.
- W3092696934 hasConceptScore W3092696934C121332964 @default.
- W3092696934 hasConceptScore W3092696934C154945302 @default.
- W3092696934 hasConceptScore W3092696934C165801399 @default.
- W3092696934 hasConceptScore W3092696934C177264268 @default.
- W3092696934 hasConceptScore W3092696934C186644900 @default.
- W3092696934 hasConceptScore W3092696934C199360897 @default.
- W3092696934 hasConceptScore W3092696934C204321447 @default.
- W3092696934 hasConceptScore W3092696934C206134035 @default.
- W3092696934 hasConceptScore W3092696934C41008148 @default.
- W3092696934 hasConceptScore W3092696934C53893814 @default.
- W3092696934 hasConceptScore W3092696934C62520636 @default.
- W3092696934 hasConceptScore W3092696934C66322947 @default.
- W3092696934 hasLocation W30926969341 @default.
- W3092696934 hasOpenAccess W3092696934 @default.
- W3092696934 hasPrimaryLocation W30926969341 @default.
- W3092696934 hasRelatedWork W2141455202 @default.
- W3092696934 hasRelatedWork W2152346074 @default.
- W3092696934 hasRelatedWork W2250398693 @default.
- W3092696934 hasRelatedWork W2251314312 @default.
- W3092696934 hasRelatedWork W2252099096 @default.
- W3092696934 hasRelatedWork W2566475580 @default.
- W3092696934 hasRelatedWork W2588225431 @default.
- W3092696934 hasRelatedWork W2612953412 @default.
- W3092696934 hasRelatedWork W2765558935 @default.
- W3092696934 hasRelatedWork W2895348981 @default.
- W3092696934 hasRelatedWork W2948992961 @default.
- W3092696934 hasRelatedWork W2970008578 @default.
- W3092696934 hasRelatedWork W2972466365 @default.
- W3092696934 hasRelatedWork W3009761736 @default.
- W3092696934 hasRelatedWork W3016560063 @default.
- W3092696934 hasRelatedWork W3022439963 @default.
- W3092696934 hasRelatedWork W3103427938 @default.
- W3092696934 hasRelatedWork W3117713602 @default.
- W3092696934 hasRelatedWork W3155000481 @default.
- W3092696934 hasRelatedWork W2379629913 @default.
- W3092696934 isParatext "false" @default.
- W3092696934 isRetracted "false" @default.
- W3092696934 magId "3092696934" @default.
- W3092696934 workType "article" @default.