Unsupervised Domain Adaptation (UDA) of the Aspect-based Sentiment Analysis (ABSA) task aims to transfer knowledge learned from labeled source domain datasets to unlabeled target domains on the assumption that samples from the source domain are freely accessible during the training period. However, this assumption can easily lead to privacy invasion issues in real-world applications, especially when the source data involves privacy-preserving domains such as healthcare and finance. In this paper, we introduce the Source-Free Domain Adaptation Framework for ABSA (SF-ABSA), which only allows model parameter transfer, not data transfer, between different domains. Specifically, the proposed SF-ABSA framework consists of two parts, i.e., feature-based adaptation and pseudo-label-based adaptation. Experiment results on four benchmarks show that the proposed framework performs competitively with traditional unsupervised domain adaptation methods under the premise of insufficient information, which demonstrates the superiority of our method under privacy conditions.

Question generation (QG) aims to automatically generate fluent and relevant questions, where the two most mainstream directions are generating questions from unstructured contextual texts (CQG), such as news articles, and generating questions from structured factoid texts (FQG), such as knowledge graphs or tables. Existing methods for these two tasks mainly face challenges of limited internal structural information as well as scarce background information, while these two tasks can benefit each other for alleviating these issues. For example, when meeting the entity mention “United Kingdom” in CQG, it can be inferred that it is a country in European continent based on the structural knowledge “(Europe, countries_within, United Kingdom)” in FQG. And when meeting the entity “Houston Rockets” in FQG, more background information, such as “an American professional basketball team based in Houston since 1971”, can be found in the related passages of CQG. To this end, we propose a unified framework for the tasks of CQG and FQG, where: (i) two types of task-sharing modules are developed to learn shared contextual and structural knowledge, where the task format is unified with a pseudo passage reformulation strategy; (ii) for the CQG task, a task-specific knowledge module with a knowledge selection and aggregation mechanism is introduced, so as to incorporate more factoid knowledge from external knowledge graphs and alleviate the word ambiguity problem; and (iii) for the FQG task, a task-specific passage module with a multi-level passage fusion mechanism is designed to extract fine-grained word-level knowledge. Experimental results in both automatic and human evaluation show the effectiveness of our proposed method.

Based on the assumption that samples in the source and target domains are freely accessible during training, unsupervised domain adaptation (UDA) of question answering (QA) aims to transfer knowledge learned from labeled source datasets to similar tasks in the unlabeled target domains. However, such assumption can easily lead to privacy violation issues in real-world applications, especially when the source domain data involves privacy-intensive domains such as finance and healthcare. In this paper, we introduce Source-Free Domain Adaptation Framework for QA (denoted as SFQA), which only allows access to trained source models for target learning, making data privacy protection more promising. Specifically, the proposed SFQA model consists of a feature extractor module (Bert Encoder) and a classifier module (Answer Classifier). We first transfer the trained source model to the target model while keeping the source classifier module frozen. Then we adopt the question generation model to generate questions and answers for the target domain. Taking the generated questions and target domain context as inputs, and the generated answers as pseudo-labels, we train the target model with joint entropy to learn a target domain-specific feature extractor. Experimental results demonstrate the superiority and effectiveness of the proposed SFQA, and show that SFQA outperforms the state-of-the-art methods.

Cross-modal retrieval has attracted much attention lately for its various applications in Internet data mining. Existing approaches mainly adopt the projection function learning paradigm to construct dual-stream models, which suffer from two limitations: 1) They only utilize the correlations provided by cross-modal data pairs but the multiple correlations among data are unexplored. 2) They typically face the challenge of abstractness of semantics, which means that an instance may have distinct semantic information in different scenarios. In this paper, we propose a novel graph learning based framework termed Multimodal Graph Learning for cross-modal retrieval (MGL), which aims to fully exploit multiple correlations embedded in multimodal data and leverage a graph neural network to capture complementary information to alleviate the information sparsity and abstractness of semantics. First, we propose a graph construction algorithm to explore diverse multimedia information. Second, a modal feature projector is designed to learn modality-shared information, and a co-attention mechanism module is proposed to capture complementary information and perform dynamic feature integration. Third, a fusion and gate module is proposed to fully aggregate captured information and perform denoising. Furthermore, we employ a graph sampling algorithm to make our approach flexible to large-scale scenarios. Experimental results on three benchmark datasets prove the effectiveness of MGL.

Knowledge graph embedding aims at modeling knowledge by projecting entities and relations into a low-dimensional semantic space. Most of the works on knowledge graph embedding construct negative samples by negative sampling as knowledge graphs typically only contain positive facts. Although substantial progress has been made by dynamic distribution based sampling methods, selecting plausible and prior information-engaged negative samples still poses many challenges. Inspired by type constraint methods, we propose Hierarchical Type Enhanced Negative Sampling (HTENS) which leverages hierarchical entity type information and entity-relation cooccurrence information to optimize the sampling probability distribution of negative samples. The experiments performed on the link prediction task demonstrate the effectiveness of HTENS. Additionally, HTENS shows its superiority in versatility and can be integrated into scalable systems with enhanced negative sampling.

Summary To identify recyclable garbage via the garbage classification is an effective countermeasure for protecting the environment. An automatic classification system supported by image recognition technologies is able to significantly reduce huge human labors of recycling tasks. However, performing the real‐time and accurate garbage detection is not a trivial task. In this article, we present an edge‐cloud framework equipped with deep learning model for recyclable garbage detection. Specifically, we propose to use the deep convolutional neural network for garbage images classification, and thus design the collaborative mechanism between edge devices and the cloud server. As a result, we design and develop a novel recyclable garbage detection system, where scanning garbage images and thus detecting recyclable ones can be completed in real‐time. We validate the performances of the proposed recyclable garbage detection system on 1000 real‐life household garbage images. Experimental results show the overall accuracy of our system reaches nearly 90%, and the time for detection is less than 500 ms.