Adaptive Semi-supervised Learning for Cross-domain Sentiment Classification

He, Ruidan; Lee, Wee Sun; Ng, Hwee Tou; Dahlmeier, Daniel

doi:10.18653/v1/d18-1383

Cited by 56 publications

(54 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Beauty, Book, Music: These data sets each consist of 6000 beauty product, book and music reviews with a balanced distribution of three class labels. This data set was introduced by (He et al, 2018). Data is obtained by sampling from a larger data set of product reviews obtained from Amazon (McAuley et al, 2015).…”

Section: Resultsmentioning

confidence: 99%

Shallow Domain Adaptive Embeddings for Sentiment Analysis

Sarma¹,

Liang²,

Sethares³

2019

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conferen

View full text Add to dashboard Cite

This paper proposes a way to improve the performance of existing algorithms for text classification in domains with strong language semantics. We propose a domain adaptation layer learns weights to combine a generic and a domain specific (DS) word embedding into a domain adapted (DA) embedding. The DA word embeddings are then used as inputs to a generic encoder + classifier framework to perform a downstream task such as classification. This adaptation layer is particularly suited to datasets that are modest in size, and which are, therefore, not ideal candidates for (re)training a deep neural network architecture. Results on binary and multi-class classification tasks using popular encoder architectures, including current state-of-the-art methods (with and without the shallow adaptation layer) show the effectiveness of the proposed approach.

show abstract

Section: Resultsmentioning

confidence: 99%

Shallow Domain Adaptive Embeddings for Sentiment Analysis

Sarma¹,

Liang²,

Sethares³

2019

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conferen

View full text Add to dashboard Cite

show abstract

“…In our implementation, the feature encoder G consists of three parts including a 300-dimensional word embedding layer using GloVe (Pennington et al, 2014), a one-layer CNN with ReLU activation function adopted in (Yu and Jiang, 2016;He et al, 2018) and a max-over-time pooling through which final sentence representation is obtained. Specifically, the convolution filter and the window size of this one-layer CNN are 300 and 3 separately.…”

Section: Training Details and Hyper-parametersmentioning

confidence: 99%

“…During training period, λ 1 , λ 2 , λ 3 , λ 4 , and n are set to 5.0, 0.1, 0.1, 1.5, 2. Similar to (He et al, 2018), we parametrize λ 4 as a dynamic weight exp[−5(1 − t max−epochs ) 2 ]λ 4 . This is to minimize the effort of the regularizer as the predictor is not good at the beginning of training.…”

Section: Training Details and Hyper-parametersmentioning

confidence: 99%

Adversarial Category Alignment Network for Cross-domain Sentiment Classification

Zou

Yang

et al. 2019

Proceedings of the 2019 Conference of the North

View full text Add to dashboard Cite

Cross-domain sentiment classification aims to predict sentiment polarity on a target domain utilizing a classifier learned from a source domain. Most existing adversarial learning methods focus on aligning the global marginal distribution by fooling a domain discriminator, without taking category-specific decision boundaries into consideration, which can lead to the mismatch of category-level features. In this work, we propose an adversarial category alignment network (ACAN), which attempts to enhance category consistency between the source domain and the target domain. Specifically, we increase the discrepancy of two polarity classifiers to provide diverse views, locating ambiguous features near the decision boundaries. Then the generator learns to create better features away from the category boundaries by minimizing this discrepancy. Experimental results on benchmark datasets show that the proposed method can achieve stateof-the-art performance and produce more discriminative features.

show abstract

“…Existing relevant studies could be attributed into two categories: two-stage approaches (Blitzer, Dredze, and Pereira 2007;Glorot, Bordes, and Bengio 2011;Ziser and Reichart 2018; and end-to-end models (Ganin et al 2016;He et al 2018;Qu et al 2019). The two-stage approaches typically construct unsupervised feature extractors or manually select pivot features across domains in the first stage.…”

Section: Introductionmentioning

confidence: 99%

“…That is, their sentiment classifiers and feature extractors overlook the sentiment polarity lying in the review text of target domains. One relevant study (He et al 2018) in this regard obtains pseudo-labels in target domains by a self-ensemble bootstrapping technique to train its sentiment classifier and feature extractor. However, the pseudo-labels are asynchronously generated by the earlier version of the sentiment classifier, which is a weaker classifier compared with its current version, and possibly limits the effectiveness of training.…”

Section: Introductionmentioning

confidence: 99%

Improving Domain-Adapted Sentiment Classification by Deep Adversarial Mutual Learning

Xue

Zhang

Zha

2020

AAAI

View full text Add to dashboard Cite

Domain-adapted sentiment classification refers to training on a labeled source domain to well infer document-level sentiment on an unlabeled target domain. Most existing relevant models involve a feature extractor and a sentiment classifier, where the feature extractor works towards learning domain-invariant features from both domains, and the sentiment classifier is trained only on the source domain to guide the feature extractor. As such, they lack a mechanism to use sentiment polarity lying in the target domain. To improve domain-adapted sentiment classification by learning sentiment from the target domain as well, we devise a novel deep adversarial mutual learning approach involving two groups of feature extractors, domain discriminators, sentiment classifiers, and label probers. The domain discriminators enable the feature extractors to obtain domain-invariant features. Meanwhile, the label prober in each group explores document sentiment polarity of the target domain through the sentiment prediction generated by the classifier in the peer group, and guides the learning of the feature extractor in its own group. The proposed approach achieves the mutual learning of the two groups in an end-to-end manner. Experiments on multiple public datasets indicate our method obtains the state-of-the-art performance, validating the effectiveness of mutual learning through label probers.

show abstract

Adaptive Semi-supervised Learning for Cross-domain Sentiment Classification

Cited by 56 publications

References 17 publications

Shallow Domain Adaptive Embeddings for Sentiment Analysis

Shallow Domain Adaptive Embeddings for Sentiment Analysis

Adversarial Category Alignment Network for Cross-domain Sentiment Classification

Improving Domain-Adapted Sentiment Classification by Deep Adversarial Mutual Learning

Contact Info

Product

Resources

About