Exploiting the Matching Information in the Support Set for Few Shot Event Classification

Lai, Viet Dac; Dernoncourt, Franck; Nguyen, Thien Huu

doi:10.1007/978-3-030-47436-2_18

Cited by 29 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, our extensive experiments on two benchmark datasets show the superior performance of our method. In the future, we will extend our analysis to other Information Extraction tasks, such as Entity Mention Detection (Nguyen et al, 2016) and Event Classification (Lai et al, 2020;.…”

Section: Discussionmentioning

confidence: 99%

A Spectral Viewpoint on Continual Relation Extraction

Nguyen,

Ngo

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

Continual Relation Extraction (CRE) aims to continuously train a model to learn new relations while preserving its ability on previously learned relations. Similar to other continual learning problems, in CRE, models experience representation shift, where learned deep space changes in the continual learning process, which leads to the downgrade in the performance of the old tasks. In this work, we will provide an insight into this phenomenon under the spectral viewpoint. Our key argument is that, for each class shape, if its eigenvectors (or spectral components) do not change much, the shape is well-preserved. We then conduct a spectral experiment and show that, for the shape of each class, the eigenvectors with larger eigenvalue are more preserved after learning new tasks which means these vectors are good at keeping class shapes. Based on this analysis, we propose a simple yet effective class-wise regularization that improve the eigenvalues in the representation learning. We observe that our proposed regularization leads to an increase in the eigenvalues. Extensive experiments on two benchmark datasets, FewRel and TACRED, show the effectiveness of our proposed method with significant improvement in performance compared to the state-of-the-art models. Further analyses also verify our hypothesis that larger eigenvalues lead to better performance and vice versa.

show abstract

Section: Discussionmentioning

confidence: 99%

A Spectral Viewpoint on Continual Relation Extraction

Nguyen,

Ngo

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

show abstract

“…The above three assumptions also apply to few-shot learning, which is a developing branch of meta-learning. Few-shot learning has been increasingly introduced into the field of NLP, such as text classification [6][7], word sense disambiguation [8], event detection [9][10], and so on.…”

Section: Meta-learningmentioning

confidence: 99%

Few shot text classification using adaptive cross capsule network

Qin

Yan²,

Chen³

2023

International Conference on Artificial Intelligence, Virtual Reality, and Visualization (AIVRV 2022)

View full text Add to dashboard Cite

In recent years, meta-learning has become a mainstream technique for few-shot learning, and it has been widely used and achieved good results in computer vision and image processing. Based on this powerful empirical performance, we are interested in using Meta-learning frameworks in NLP to deal with the task of few-shot learning (FSL). However, due to the sparse sample size, sample-level comparisons based on other expressions are highly susceptible to interference, leading to serious overfitting problems. To achieve classification tasks, we suggest a novel Adaptive Cross-Capsule Network (ACCN) for learning generalized representations. A dynamic routing technique is utilized with the concept of a prototype network to train the support set to generalize the generalized representations of each category. The support set and the query set can fully interact dynamically to capture the essential semantic aspects of the query set following a successful non-parametric cross-attention method. Experimental results show that ACCN proposed in this paper is well adaptive to the intention classification task under additional categories, which obtain SOTA results on FewRel Datasets, which also can perform significantly better than the original classification system on Huffpost Datasets. This provides a crucial foundation for this study.

show abstract

“…A large body of previous ED research is dedicated to monolingual learning, i.e., training and testing over the same languages (Nguyen et al, 2016;Yang and Mitchell, 2016;Lu and Nguyen, 2018;Lai et al, 2020;Lin et al, 2020). The models might consider different domains Man Duc Trong et al, 2020).…”

Section: Related Workmentioning

confidence: 99%

Transitioning Representations between Languages for Cross-lingual Event Detection via Langevin Dynamics

Nguyen,

Dernoncourt

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

Cross-lingual transfer learning (CLTL) for event detection (ED) aims to develop models in high-resource source languages that can be directly applied to produce effective performance for lower-resource target languages. Previous research in this area has focused on representation matching methods to develop a language-universal representation space into which source-and target-language example representations can be mapped to achieve crosslingual transfer. However, as this approach modifies the representations for the sourcelanguage examples, the models might lose discriminative features for ED that are learned over training data of the source language to prevent effective predictions. To this end, our work introduces a novel approach for cross-lingual ED where we only aim to transition the representations for the target-language examples into the source-language space, thus preserving the representations in the source language and their discriminative information. Our method introduces Langevin Dynamics to perform representation transition and a semantic preservation framework to retain event type features during the transition process. Extensive experiments over three languages demonstrate the state-of-the-art performance for ED in CLTL.

show abstract

Exploiting the Matching Information in the Support Set for Few Shot Event Classification

Cited by 29 publications

References 18 publications

A Spectral Viewpoint on Continual Relation Extraction

A Spectral Viewpoint on Continual Relation Extraction

Few shot text classification using adaptive cross capsule network

Transitioning Representations between Languages for Cross-lingual Event Detection via Langevin Dynamics

Contact Info

Product

Resources

About