Feng-Chi Liu

Anxiety and depression have become major global health concerns. With the rapid rise of social media, people increasingly share emotions and personal struggles through posts, which often convey multiple mental states simultaneously. To address this multi-label classification challenge in mental health texts, this study proposes a multi-task framework with two main modules, a multi-perspective prompt design module and a perturbation-based self-supervised learning module, based on a pre-trained language model backbone. Prompts from sociological, psychological, and educational perspectives are used to enhance semantic understanding. To improve model robustness, we formulate self-supervised auxiliary tasks where the model predicts whether a sentence has undergone insertion, swap, or deletion. Experiments on the MultiWD dataset, covering six wellness dimensions, show that our method outperforms all baselines. Furthermore, ablation studies explore the impact of different training configurations and confirm the critical contributions of both proposed modules.

In the domain of knowledge graph embedding, conventional approaches typically transform entities and relations into continuous vector spaces. However, parameter efficiency becomes increasingly crucial when dealing with large-scale knowledge graphs that contain vast numbers of entities and relations. In particular, resource-intensive embeddings often lead to increased computational costs, and may limit scalability and adaptability in practical environments, such as in low-resource settings or real-world applications. This paper explores an approach to knowledge graph representation learning that leverages small, reserved entities and relation sets for parameter-efficient embedding. We introduce a hierarchical attention network designed to refine and maximize the representational quality of embeddings by selectively focusing on these reserved sets, thereby reducing model complexity. Empirical assessments validate that our model achieves high performance on the benchmark dataset with fewer parameters and smaller embedding dimensions. The ablation studies further highlight the impact and contribution of each component in the proposed hierarchical attention structure.

This research models and forecasts bounded ordinal time series data that can appear in various contexts, such as air quality index (AQI) levels, economic situations, and credit ratings. This class of time series data is characterized by being bounded and exhibiting a concentration of large probabilities on a few categories, such as states 0 and 1. We propose using Bayesian methods for modeling and forecasting in zero-one-inflated bounded Poisson autoregressive (ZOBPAR) models, which are specifically designed to capture the dynamic changes in such ordinal time series data. We innovatively extend models to incorporate exogenous variables, marking a new direction in Bayesian inferences and forecasting. Simulation studies demonstrate that the proposed methods accurately estimate all unknown parameters, and the posterior means of parameter estimates are robustly close to the actual values as the sample size increases. In the empirical study we investigate three datasets of daily AQI levels from three stations in Taiwan and consider five competing models for the real examples. The results exhibit that the proposed method reasonably predicts the AQI levels in the testing period, especially for the Miaoli station.

This study assesses governments' long-term non-pharmaceutical interventions upon the coronavirus disease 2019 (COVID-19) pandemic in East Asia. It advances the literature towards a better understanding of when and which control measures are effective. We (1) provide time-varying case fatality ratios and focus on the elderly's mortality and case fatality ratios, (2) measure the correlations between daily new cases (daily new deaths) and each index based on multiple domestic pandemic waves and (3) examine the lead–lag relationship between daily new cases (daily new deaths) and each index via the cross-correlation functions on the pre-whitened series. Our results show that the interventions reduce COVID-19 infections for some periods before the period of the Omicron variant. Moreover, there is no COVID-19 policy lag in Taiwan between daily new confirmed cases and each index. As of March 2022, the case fatality ratios of the elderly group in Japan, Hong Kong and South Korea are 4.69%, 4.72% and 1.48%, respectively, while the case fatality ratio of the elderly group in Taiwan is 25.01%. A government's COVID-19 vaccination distribution and prioritisation policies are pivotal for the elderly group to reduce the number of deaths. Immunising this specific group as best as possible should undoubtedly be a top priority.