Deep Semantic Role Labeling With Self-Attention

Tan, Zhixing; Wang, Mingxuan; Xie, Jun; Chen, Yidong; Shi, Xiaodong

doi:10.1609/aaai.v32i1.11928

Cited by 227 publications

(33 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Semantic Extractor: We adopt the bidirectional long shortterm memory (Bi-LSTM) (Zhou et al 2016) with selfattention mechanism (Tan et al 2018) to learn the document representation. With the input word w q in a document x, the hidden states of Bi-LSTM are updated as,…”

Section: Learning Head Label Classifiersmentioning

confidence: 99%

Does Head Label Help for Long-Tailed Multi-Label Text Classification

Xiao

Zhang

Jing

et al. 2021

AAAI

View full text Add to dashboard Cite

Multi-label text classification (MLTC) aims to annotate documents with the most relevant labels from a number of candidate labels. In real applications, the distribution of label frequency often exhibits a long tail, i.e., a few labels are associated with a large number of documents (a.k.a. head labels), while a large fraction of labels are associated with a small number of documents (a.k.a. tail labels). To address the challenge of insufficient training data on tail label classification, we propose a Head-to-Tail Network (HTTN) to transfer the meta-knowledge from the data-rich head labels to data-poor tail labels. The meta-knowledge is the mapping from few-shot network parameters to many-shot network parameters, which aims to promote the generalizability of tail classifiers. Extensive experimental results on three benchmark datasets demonstrate that HTTN consistently outperforms the state-of-the-art methods. The code and hyper-parameter settings are released for reproducibility.

show abstract

Section: Learning Head Label Classifiersmentioning

confidence: 99%

Does Head Label Help for Long-Tailed Multi-Label Text Classification

Xiao

Zhang

Jing

et al. 2021

AAAI

View full text Add to dashboard Cite

show abstract

“…Cheng et al integrated the long short-term memory architecture with self-attention mechanism to render sequence-level networks better at handling structured input [20]. Tan et al presented a simple and effective architecture for semantic role labeling based on self-attention to handle structural information and long range dependencies [21]. The attention mechanism has enjoyed great popularity in the machine translation as well as NLP communities.…”

Section: B Self-attention Mechanismmentioning

confidence: 99%

A Novel Clothing Attribute Representation Network-Based Self-Attention Mechanism

Chun

Wang

2020

IEEE Access

View full text Add to dashboard Cite

As highly increasing of on-line fashions retail industry, automatic recognition and representation of clothing items have huge potentials. With the help of deep learning methods, many clothing attribute representation models have been proposed. However, these models are mainly suitable for coarse-grained classification which are not suitable for clothing attribute representation. To address such a problem, in this article, we propose a novel network structure named SAC, which is a combination of CNNs and Self-attention mechanism and can represent clothing attributes more fine-grained. Besides, we use Grad-CAM to visualize which part of the clothing attributes is more concerned by customers. Finally, a new labeled clothing dataset is introduced in this article, which is expected to be helpful to the researchers who are working in fashion domains for image representation.

show abstract

“…Two main classes of approaches to span-based SRL are sequence labeling approaches (Yang and Mitchell 2017;He et al 2017;Tan et al 2018;Strubell et al 2018) and graphbased approaches (He et al 2018;Li et al 2019;Ouchi, Shindo, and Matsumoto 2018;Marcheggiani and Titov 2020). Sequence labeling approaches often employ the BIO tagging scheme and predict a BIO tag for each word in the input sentence.…”

Section: Introductionmentioning

confidence: 99%

Span-Based Semantic Role Labeling with Argument Pruning and Second-Order Inference

Jia¹,

Zhang²,

Wu³

et al. 2022

AAAI

View full text Add to dashboard Cite

We study graph-based approaches to span-based semantic role labeling. This task is difficult due to the need to enumerate all possible predicate-argument pairs and the high degree of imbalance between positive and negative samples. Based on these difficulties, high-order inference that considers interactions between multiple arguments and predicates is often deemed beneficial but has rarely been used in span-based semantic role labeling. Because even for second-order inference, there are already O(n^5) parts for a sentence of length n, and exact high-order inference is intractable. In this paper, we propose a framework consisting of two networks: a predicate-agnostic argument pruning network that reduces the number of candidate arguments to O(n), and a semantic role labeling network with an optional second-order decoder that is unfolded from an approximate inference algorithm. Our experiments show that our framework achieves significant and consistent improvement over previous approaches.

show abstract

Deep Semantic Role Labeling With Self-Attention

Cited by 227 publications

References 24 publications

Does Head Label Help for Long-Tailed Multi-Label Text Classification

Does Head Label Help for Long-Tailed Multi-Label Text Classification

A Novel Clothing Attribute Representation Network-Based Self-Attention Mechanism

Span-Based Semantic Role Labeling with Argument Pruning and Second-Order Inference

Contact Info

Product

Resources

About