Matches in SemOpenAlex for { <https://semopenalex.org/work/W3187364184> ?p ?o ?g. }
- W3187364184 abstract "In neural machine translation (NMT), we search for the mode of the model distribution to form predictions. The mode as well as other high probability translations found by beam search have been shown to often be inadequate in a number of ways. This prevents practitioners from improving translation quality through better search, as these idiosyncratic translations end up being selected by the decoding algorithm, a problem known as the beam search curse. Recently, a sampling-based approximation to minimum Bayes risk (MBR) decoding has been proposed as an alternative decision rule for NMT that would likely not suffer from the same problems. We analyse this approximation and establish that it has no equivalent to the beam search curse, i.e. better search always leads to better translations. We also design different approximations aimed at decoupling the cost of exploration from the cost of robust estimation of expected utility. This allows for exploration of much larger hypothesis spaces, which we show to be beneficial. We also show that it can be beneficial to make use of strategies like beam search and nucleus sampling to construct hypothesis spaces efficiently. We show on three language pairs (English into and from German, Romanian, and Nepali) that MBR can improve upon beam search with moderate computation." @default.
- W3187364184 created "2021-08-16" @default.
- W3187364184 creator A5059243935 @default.
- W3187364184 creator A5074142241 @default.
- W3187364184 date "2021-08-10" @default.
- W3187364184 modified "2023-09-28" @default.
- W3187364184 title "Sampling-Based Minimum Bayes Risk Decoding for Neural Machine Translation." @default.
- W3187364184 cites W1828163288 @default.
- W3187364184 cites W1967573895 @default.
- W3187364184 cites W1969483458 @default.
- W3187364184 cites W1985093013 @default.
- W3187364184 cites W1988520084 @default.
- W3187364184 cites W1993666847 @default.
- W3187364184 cites W2101105183 @default.
- W3187364184 cites W2108984739 @default.
- W3187364184 cites W2115526192 @default.
- W3187364184 cites W2116492146 @default.
- W3187364184 cites W2126270798 @default.
- W3187364184 cites W2136657878 @default.
- W3187364184 cites W2138414624 @default.
- W3187364184 cites W2156953672 @default.
- W3187364184 cites W2157365695 @default.
- W3187364184 cites W2158658733 @default.
- W3187364184 cites W2159358338 @default.
- W3187364184 cites W2164732909 @default.
- W3187364184 cites W2222235228 @default.
- W3187364184 cites W2250342921 @default.
- W3187364184 cites W2251367463 @default.
- W3187364184 cites W2251610689 @default.
- W3187364184 cites W2305436255 @default.
- W3187364184 cites W2508316494 @default.
- W3187364184 cites W2512924740 @default.
- W3187364184 cites W2525778437 @default.
- W3187364184 cites W2552839021 @default.
- W3187364184 cites W2566623769 @default.
- W3187364184 cites W2608146174 @default.
- W3187364184 cites W2758950307 @default.
- W3187364184 cites W2903193068 @default.
- W3187364184 cites W2909737760 @default.
- W3187364184 cites W2916548775 @default.
- W3187364184 cites W2918914336 @default.
- W3187364184 cites W2949888546 @default.
- W3187364184 cites W2950858167 @default.
- W3187364184 cites W2963248296 @default.
- W3187364184 cites W2963366552 @default.
- W3187364184 cites W2963382396 @default.
- W3187364184 cites W2963403868 @default.
- W3187364184 cites W2963506925 @default.
- W3187364184 cites W2963532001 @default.
- W3187364184 cites W2963620441 @default.
- W3187364184 cites W2963653811 @default.
- W3187364184 cites W2963919854 @default.
- W3187364184 cites W2964103964 @default.
- W3187364184 cites W2964199361 @default.
- W3187364184 cites W2964308564 @default.
- W3187364184 cites W2970646865 @default.
- W3187364184 cites W2970692082 @default.
- W3187364184 cites W2996287690 @default.
- W3187364184 cites W3035214886 @default.
- W3187364184 cites W3045998699 @default.
- W3187364184 cites W3089025530 @default.
- W3187364184 cites W3091784310 @default.
- W3187364184 cites W3105718208 @default.
- W3187364184 cites W3118026775 @default.
- W3187364184 cites W3154984521 @default.
- W3187364184 cites W3157748223 @default.
- W3187364184 cites W3172048174 @default.
- W3187364184 cites W3176846585 @default.
- W3187364184 cites W2250960671 @default.
- W3187364184 hasPublicationYear "2021" @default.
- W3187364184 type Work @default.
- W3187364184 sameAs 3187364184 @default.
- W3187364184 citedByCount "0" @default.
- W3187364184 crossrefType "posted-content" @default.
- W3187364184 hasAuthorship W3187364184A5059243935 @default.
- W3187364184 hasAuthorship W3187364184A5074142241 @default.
- W3187364184 hasConcept C106131492 @default.
- W3187364184 hasConcept C107673813 @default.
- W3187364184 hasConcept C11413529 @default.
- W3187364184 hasConcept C119857082 @default.
- W3187364184 hasConcept C125583679 @default.
- W3187364184 hasConcept C126255220 @default.
- W3187364184 hasConcept C140779682 @default.
- W3187364184 hasConcept C154945302 @default.
- W3187364184 hasConcept C19889080 @default.
- W3187364184 hasConcept C203005215 @default.
- W3187364184 hasConcept C207201462 @default.
- W3187364184 hasConcept C31972630 @default.
- W3187364184 hasConcept C33923547 @default.
- W3187364184 hasConcept C41008148 @default.
- W3187364184 hasConcept C57273362 @default.
- W3187364184 hasConceptScore W3187364184C106131492 @default.
- W3187364184 hasConceptScore W3187364184C107673813 @default.
- W3187364184 hasConceptScore W3187364184C11413529 @default.
- W3187364184 hasConceptScore W3187364184C119857082 @default.
- W3187364184 hasConceptScore W3187364184C125583679 @default.
- W3187364184 hasConceptScore W3187364184C126255220 @default.
- W3187364184 hasConceptScore W3187364184C140779682 @default.
- W3187364184 hasConceptScore W3187364184C154945302 @default.
- W3187364184 hasConceptScore W3187364184C19889080 @default.