Preliminary Conference Agenda

To prevent other types of mental health problems from being misclassified as depression, as well as to remedy the problem of inadequate resources for mental health consultations. This study first analyzes the types of different causes of mental health problems, providing an important basis for better understanding the diversity and complexity of this field. Subsequently, a machine learning approach was used to predict the potential causes of dif-ferent types of mental health problems. This research provides new perspec-tives and methods for early identification and personalized treatment of mental health problems. The experimental results show that depression ac-counts for only 16.9% of mental health problems. In the prediction of the causes of mental health problems, the SVM method performed best in pre-dicting the causes of mental health problems, outperforming 5 machine learning methods and 3 deep learning methods. Through these studies, we hope to prevent other types of mental health problems from being misclassi-fied as depression and to remedy the lack of resources for mental health counseling. This will help increase the success rate of early intervention and provide better mental health support for patients.

This paper describes the developing process and preliminary testing of a physical representation system aimed at supporting preschool children's picture book search. The authors first conducted named entity recognition (NER) on a corpus of 880 picture book summaries to supplement a metadata schema identified in prior research. They then designed a system using colored stripes and icons to physically represent these metadata elements on book spines. A small-scale experiment (N=8) comparing search performance between children taught the representation system versus untaught controls was conducted. The results suggest that the representation system can be understood by children and improves recall, precision, and success rates. The findings provide initial evidence that mapping metadata to intuitive physical identifiers could enhance young children's book search experiences and engagement. Further research with larger samples is needed to evaluate the effectiveness of this approach.

Abstract. Billions of scientific papers lead to the need to identify essential parts from the massive text. Scientific research is an activity from putting forward problems to using methods. To learn the main idea from scientific papers, we fo-cus on extracting problem and method sentences. Annotating sentences in scien-tific papers is labor-intensive, resulting in the creation of small-scale datasets that limit model learning. To tackle this challenge, data augmentation has been adopted due to its ability to generate synthetic data with minor variations, thereby expand-ing the scale of the original training dataset. Nowadays, there are various data augmentation methods, such as those based on random word replacement or back translation. Nevertheless, their suitability for sentence classification tasks in sci-entific papers remains unexplored. Thus, this paper constructs two manually an-notation datasets and evaluates their performance. Furthermore, this paper delves into the mechanisms underlying their effects. Previous studies have suggested that data augmentation can diminish reliance on high-frequency patterns in mod-els. Therefore, this paper employs attention values to represent the model's de-pendence on words, and analyzes how data augmentation methods alter the atten-tion values of individual words within sentences. The experimental results indi-cate that data augmentation methods can improve the macro F1 score in sentence classification tasks. Furthermore, data augmentation methods effectively reduce the attention values assigned to stop words, commonly used words in scientific papers, and commonly used words in method and problem sentences.