Panfeng Cao, Ye Wang, Qiang Zhang, Zaiqiao Meng. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.

Generative adversarial networks as a powerful technique is also used in domain adaptation (DA) problem. Existing adversarial DA methods mainly conduct domain-wise alignment to alleviate marginal distribution shift between the two domains, while it may damage latent discriminative structure hidden in data feature space and cause negative transfer accordingly. To handle this problem, we propose a joint adversarial domain adaptation method with structural graph alignment to minimize joint distribution bias by further realizing class-wise matching (conditional distribution shift) based on a simple sampling strategy except for the domain-wise alignment, and validate that simultaneously considering these two types of shift can approximately reduce the joint distribution bias. To explore transferable structural information and realize more sufficient transfer for DA problem, we propose to align structural graphs between the two domains which is also based on a simple sampling strategy. Notably, the structural graph describes the relationship between each two samples and it is computed on two domains. As such, we can learn new feature representation of the two domains which are more discriminative and transferable to benefit a cross-domain classification task desirably. Finally, we design a number of experiments to evaluate our approach on four public cross-domain benchmark datasets including standard and large-scale ones, and empirical results show that the proposed model can outperform compared state-of-the-art methods.

The task of domain generalization is to learn a classification model from multiple source domains and generalize it to unknown target domains. The key to domain generalization is learning discriminative domain-invariant features. Invariant representations are achieved using adversarial domain generalization as one of the primary techniques. For example, generative adversarial networks have been widely used, but suffer from the problem of low intra-class diversity, which can lead to poor generalization ability. To address this issue, we propose a new method called auxiliary classifier in adversarial domain generalization (CloCls). CloCls improve the diversity of the source domain by introducing auxiliary classifier. Combining typical task-related losses, e.g., cross-entropy loss for classification and adversarial loss for domain discrimination, our overall goal is to guarantee the learning of condition-invariant features for all source domains while increasing the diversity of source domains. Further, inspired by smoothing optima have improved generalization for supervised learning tasks like classification. We leverage that converging to a smooth minima with respect task loss stabilizes the adversarial training leading to better performance on unseen target domain which can effectively enhances the performance of domain adversarial methods. We have conducted extensive image classification experiments on benchmark datasets in domain generalization, and our model exhibits sufficient generalization ability and outperforms state-of-the-art DG methods.

Emotional tracking on time-varying virtual space communication aims to identify sentiments and opinions expressed in a piece of user-generated content. However, the existing research mainly focuses on the user’s single post, despite the fact that social network data are sequential. In this article, we propose a sentiment analysis model based on time series prediction in order to understand and master the chronological evolution of the user’s point of view. Specifically, with the help of a domain-knowledge-enhanced pre-trained encoder, the model embeds tokens for each moment in the text sequence. We then propose an attention-based temporal prediction model to extract rich timing information from historical posting records, which improves the prediction of the user’s current state and personalizes the analysis of user’s sentiment changes in social networks. The experiments show that the proposed model improves on four kinds of sentiment tasks and significantly outperforms the strong baseline.

With the increasing volume of healthcare data, automated International Classification of Diseases (ICD) has become increasingly relevant and is frequently regarded as a medical multi-label prediction problem. Current methods struggle to accurately classify medical diagnosis texts that represent deep and sparse categories. Unlike these works that model the label with code hierarchy or description for label prediction, we argue that the label generation with structural information can provide more comprehensive knowledge based on the observation that label synonyms and parent-child relationships in vary from their context in clinical contexts. In this study, we introduce \tool, a heterogeneous graph model with improved attention for automated ICD coding. Notably, our approach represents the model to consider this task as a labelled graph generation problem. Our enhanced attention mechanism boosts the model's capacity to learn from multi-relational heterogeneous graph representations. Additionally, we propose a discriminator for labelled graphs (LG) that computes the reward for each ICD code in the labelled graph generator. Our experimental findings demonstrate that our proposed model significantly outperforms all existing strong baseline methods and attains the best performance on three benchmark datasets.

Multidocument aspect-based summarization (AspSumm) aims to generate focused summaries based on the target aspects from a cluster of relevant documents. Generating such summaries can better satisfy readers’ specific points of interest, as readers may have different concerns about the same articles. However, previous methods usually generate aspect-based summaries based on the given aspects without using the relationship among aspects to assist in the summarization. In this work, we propose a two-stage general framework for multidocument AspSumm. The model first discovers the latent relationship among aspects and then uses relevant sentences selected by aspect discovery to generate abstractive summaries. We exploit latent dependencies among aspects using a tag mask training (TMT) strategy, which increases the interpretability of the model. In addition to improvements in summarization over aspect-based strong baselines, experimental results show that our proposed model can accurately discover multidomain aspects on the WikiAsp dataset.