Matches in SemOpenAlex for { <https://semopenalex.org/work/W3210031772> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3210031772 abstract "Aspect-based Sentiment Analysis (ABSA) aims to determine the sentiment polarity towards an aspect. Because of the expensive and limited labelled data, the pretraining strategy has become the de-facto standard for ABSA. However, there always exists severe domain shift between the pretraining and downstream ABSA datasets, hindering the effective knowledge transfer when directly finetuning and making the downstream task performs sub-optimal. To mitigate such domain shift, we introduce a unified alignment pretraining framework into the vanilla pretrain-finetune pipeline with both instance- and knowledge-level alignments. Specifically, we first devise a novel coarse-to-fine retrieval sampling approach to select target domain-related instances from the large-scale pretraining dataset, thus aligning the instances between pretraining and target domains (textit{First Stage}). Then, we introduce a knowledge guidance-based strategy to further bridge the domain gap at the knowledge level. In practice, we formulate the model pretrained on the sampled instances into a knowledge guidance model and a learner model, respectively. On the target dataset, we design an on-the-fly teacher-student joint fine-tuning approach to progressively transfer the knowledge from the knowledge guidance model to the learner model (textit{Second Stage}). Thereby, the learner model can maintain more domain-invariant knowledge when learning new knowledge from the target dataset. In the textit{Third Stage,} the learner model is finetuned to better adapt its learned knowledge to the target dataset. Extensive experiments and analyses on several ABSA benchmarks demonstrate the effectiveness and universality of our proposed pretraining framework. Notably, our pretraining framework pushes several strong baseline models up to the new state-of-the-art records. We release our code and models." @default.
- W3210031772 created "2021-11-08" @default.
- W3210031772 creator A5001819736 @default.
- W3210031772 creator A5004758964 @default.
- W3210031772 creator A5031572143 @default.
- W3210031772 creator A5039173319 @default.
- W3210031772 creator A5053183023 @default.
- W3210031772 creator A5074700770 @default.
- W3210031772 date "2021-10-26" @default.
- W3210031772 modified "2023-10-17" @default.
- W3210031772 title "Unified Instance and Knowledge Alignment Pretraining for Aspect-based Sentiment Analysis" @default.
- W3210031772 cites W1832693441 @default.
- W3210031772 cites W2146502635 @default.
- W3210031772 cites W2153579005 @default.
- W3210031772 cites W2155482025 @default.
- W3210031772 cites W2160660844 @default.
- W3210031772 cites W2250539671 @default.
- W3210031772 cites W2251124635 @default.
- W3210031772 cites W2251648804 @default.
- W3210031772 cites W2562607067 @default.
- W3210031772 cites W2592691248 @default.
- W3210031772 cites W2740899359 @default.
- W3210031772 cites W2757016771 @default.
- W3210031772 cites W2788810909 @default.
- W3210031772 cites W2799009183 @default.
- W3210031772 cites W2891300059 @default.
- W3210031772 cites W2903110172 @default.
- W3210031772 cites W2949161734 @default.
- W3210031772 cites W2950404230 @default.
- W3210031772 cites W2962808042 @default.
- W3210031772 cites W2963012544 @default.
- W3210031772 cites W2963026768 @default.
- W3210031772 cites W2963240575 @default.
- W3210031772 cites W2963341956 @default.
- W3210031772 cites W2963428430 @default.
- W3210031772 cites W2964098749 @default.
- W3210031772 cites W2971220558 @default.
- W3210031772 cites W2991433488 @default.
- W3210031772 cites W3029506370 @default.
- W3210031772 cites W3034238904 @default.
- W3210031772 cites W3035529900 @default.
- W3210031772 cites W3044187822 @default.
- W3210031772 cites W3100060077 @default.
- W3210031772 cites W3100456868 @default.
- W3210031772 cites W3104215796 @default.
- W3210031772 cites W3106996681 @default.
- W3210031772 cites W3167830325 @default.
- W3210031772 cites W3173982660 @default.
- W3210031772 cites W3176719207 @default.
- W3210031772 cites W66373487 @default.
- W3210031772 doi "https://doi.org/10.48550/arxiv.2110.13398" @default.
- W3210031772 hasPublicationYear "2021" @default.
- W3210031772 type Work @default.
- W3210031772 sameAs 3210031772 @default.
- W3210031772 citedByCount "0" @default.
- W3210031772 crossrefType "posted-content" @default.
- W3210031772 hasAuthorship W3210031772A5001819736 @default.
- W3210031772 hasAuthorship W3210031772A5004758964 @default.
- W3210031772 hasAuthorship W3210031772A5031572143 @default.
- W3210031772 hasAuthorship W3210031772A5039173319 @default.
- W3210031772 hasAuthorship W3210031772A5053183023 @default.
- W3210031772 hasAuthorship W3210031772A5074700770 @default.
- W3210031772 hasBestOaLocation W32100317721 @default.
- W3210031772 hasConcept C119857082 @default.
- W3210031772 hasConcept C134306372 @default.
- W3210031772 hasConcept C154945302 @default.
- W3210031772 hasConcept C204321447 @default.
- W3210031772 hasConcept C207685749 @default.
- W3210031772 hasConcept C2776960227 @default.
- W3210031772 hasConcept C33923547 @default.
- W3210031772 hasConcept C36503486 @default.
- W3210031772 hasConcept C41008148 @default.
- W3210031772 hasConcept C56739046 @default.
- W3210031772 hasConceptScore W3210031772C119857082 @default.
- W3210031772 hasConceptScore W3210031772C134306372 @default.
- W3210031772 hasConceptScore W3210031772C154945302 @default.
- W3210031772 hasConceptScore W3210031772C204321447 @default.
- W3210031772 hasConceptScore W3210031772C207685749 @default.
- W3210031772 hasConceptScore W3210031772C2776960227 @default.
- W3210031772 hasConceptScore W3210031772C33923547 @default.
- W3210031772 hasConceptScore W3210031772C36503486 @default.
- W3210031772 hasConceptScore W3210031772C41008148 @default.
- W3210031772 hasConceptScore W3210031772C56739046 @default.
- W3210031772 hasLocation W32100317721 @default.
- W3210031772 hasOpenAccess W3210031772 @default.
- W3210031772 hasPrimaryLocation W32100317721 @default.
- W3210031772 hasRelatedWork W1493504683 @default.
- W3210031772 hasRelatedWork W1525380347 @default.
- W3210031772 hasRelatedWork W1525691822 @default.
- W3210031772 hasRelatedWork W1551489865 @default.
- W3210031772 hasRelatedWork W2568140311 @default.
- W3210031772 hasRelatedWork W2886017866 @default.
- W3210031772 hasRelatedWork W2911810434 @default.
- W3210031772 hasRelatedWork W2968586400 @default.
- W3210031772 hasRelatedWork W3175655865 @default.
- W3210031772 hasRelatedWork W4210794429 @default.
- W3210031772 isParatext "false" @default.
- W3210031772 isRetracted "false" @default.
- W3210031772 magId "3210031772" @default.
- W3210031772 workType "article" @default.