Matches in SemOpenAlex for { <https://semopenalex.org/work/W2913309424> ?p ?o ?g. }
Showing items 1 to 85 of
85
with 100 items per page.
- W2913309424 abstract "The cross entropy (CE) loss function is commonly adopted for neural network language model (NNLM) training. Although this criterion is largely successful, as evidenced by the quick advance of NNLM, minimizing CE only maximizes likelihood of training data. When training data is insufficient, the generalization power of the resulting LM is limited on test data. In this paper, we propose to integrate a pairwise ranking (PR) loss with the CE loss for multi-objective training on recurrent neural network language model (RNNLM). The PR loss emphasizes discrimination between target and non-target words and also reserves probabilities for low-frequency correct words, which complements the distribution learning role of the CE loss. Combining the two losses may therefore help improve the performance of RNNLM. In addition, we incorporate multi-task learning (MTL) into the proposed multi-objective learning to regularize the primary task of RNNLM by an auxiliary task of part-of-speech (POS) tagging. The proposed approach to RNNLM learning has been evaluated on two speech recognition tasks of WSJ and AMI with encouraging results achieved on word error rate reductions." @default.
- W2913309424 created "2019-02-21" @default.
- W2913309424 creator A5028632429 @default.
- W2913309424 creator A5033518479 @default.
- W2913309424 creator A5049905128 @default.
- W2913309424 date "2018-12-01" @default.
- W2913309424 modified "2023-10-16" @default.
- W2913309424 title "Multi-Objective Multi-Task Learning on RNNLM for Speech Recognition" @default.
- W2913309424 cites W1566877644 @default.
- W2913309424 cites W179875071 @default.
- W2913309424 cites W1986767396 @default.
- W2913309424 cites W1989705153 @default.
- W2913309424 cites W1996430422 @default.
- W2913309424 cites W2024490156 @default.
- W2913309424 cites W2057653135 @default.
- W2913309424 cites W2058695628 @default.
- W2913309424 cites W2064675550 @default.
- W2913309424 cites W2104032746 @default.
- W2913309424 cites W2110415041 @default.
- W2913309424 cites W2134731454 @default.
- W2913309424 cites W2158148237 @default.
- W2913309424 cites W2164628107 @default.
- W2913309424 cites W2171928131 @default.
- W2913309424 cites W2396464458 @default.
- W2913309424 cites W2401969231 @default.
- W2913309424 cites W2749738822 @default.
- W2913309424 cites W4230318854 @default.
- W2913309424 doi "https://doi.org/10.1109/slt.2018.8639649" @default.
- W2913309424 hasPublicationYear "2018" @default.
- W2913309424 type Work @default.
- W2913309424 sameAs 2913309424 @default.
- W2913309424 citedByCount "2" @default.
- W2913309424 countsByYear W29133094242021 @default.
- W2913309424 countsByYear W29133094242022 @default.
- W2913309424 crossrefType "proceedings-article" @default.
- W2913309424 hasAuthorship W2913309424A5028632429 @default.
- W2913309424 hasAuthorship W2913309424A5033518479 @default.
- W2913309424 hasAuthorship W2913309424A5049905128 @default.
- W2913309424 hasConcept C119857082 @default.
- W2913309424 hasConcept C137293760 @default.
- W2913309424 hasConcept C147168706 @default.
- W2913309424 hasConcept C153180895 @default.
- W2913309424 hasConcept C154945302 @default.
- W2913309424 hasConcept C162324750 @default.
- W2913309424 hasConcept C167981619 @default.
- W2913309424 hasConcept C184898388 @default.
- W2913309424 hasConcept C187736073 @default.
- W2913309424 hasConcept C2780451532 @default.
- W2913309424 hasConcept C28006648 @default.
- W2913309424 hasConcept C28490314 @default.
- W2913309424 hasConcept C40969351 @default.
- W2913309424 hasConcept C41008148 @default.
- W2913309424 hasConcept C50644808 @default.
- W2913309424 hasConceptScore W2913309424C119857082 @default.
- W2913309424 hasConceptScore W2913309424C137293760 @default.
- W2913309424 hasConceptScore W2913309424C147168706 @default.
- W2913309424 hasConceptScore W2913309424C153180895 @default.
- W2913309424 hasConceptScore W2913309424C154945302 @default.
- W2913309424 hasConceptScore W2913309424C162324750 @default.
- W2913309424 hasConceptScore W2913309424C167981619 @default.
- W2913309424 hasConceptScore W2913309424C184898388 @default.
- W2913309424 hasConceptScore W2913309424C187736073 @default.
- W2913309424 hasConceptScore W2913309424C2780451532 @default.
- W2913309424 hasConceptScore W2913309424C28006648 @default.
- W2913309424 hasConceptScore W2913309424C28490314 @default.
- W2913309424 hasConceptScore W2913309424C40969351 @default.
- W2913309424 hasConceptScore W2913309424C41008148 @default.
- W2913309424 hasConceptScore W2913309424C50644808 @default.
- W2913309424 hasLocation W29133094241 @default.
- W2913309424 hasOpenAccess W2913309424 @default.
- W2913309424 hasPrimaryLocation W29133094241 @default.
- W2913309424 hasRelatedWork W179875071 @default.
- W2913309424 hasRelatedWork W1984537751 @default.
- W2913309424 hasRelatedWork W2396464458 @default.
- W2913309424 hasRelatedWork W2401653770 @default.
- W2913309424 hasRelatedWork W2507865718 @default.
- W2913309424 hasRelatedWork W2594897229 @default.
- W2913309424 hasRelatedWork W2810852753 @default.
- W2913309424 hasRelatedWork W2913309424 @default.
- W2913309424 hasRelatedWork W2963308316 @default.
- W2913309424 hasRelatedWork W3025310782 @default.
- W2913309424 isParatext "false" @default.
- W2913309424 isRetracted "false" @default.
- W2913309424 magId "2913309424" @default.
- W2913309424 workType "article" @default.