International Journal of Smart Computing and Artificial Intelligence

Simile Identification with Pseudo Data Acquisition and Re-labeling

2024-06-04T14:34:46+00:00

The simile is a kind of figurative language. It expresses the target of the figurative language by using some typical phrases such as “like”. It is important to distinguish whether the sentence is a simile or a literal for understanding a sentence. However, a large amount of data is required to generate a classifier by machine learning. Moreover, creating the dataset is costly. In this paper, we propose a pseudo dataset acquisition method for simile identification. We first construct a dataset of simile and literal sentences using machine translation. We utilize mBART as the machine translation system. This process automatically generates pseudo-simile and literal instances from three types of corpora. Then, we apply some machine learning approaches to the simile identification task. We compare Support Vector Machine, Naive Bayes, and BERT in the experiment. The experimental result shows the validity of the pseudo dataset as compared with a simple baseline (machine translation with rules). In addition, re-labeling with machine learning for the original pseudo data contributed to the improvement of the simile identification accuracy.

Self-mental Care Management System by Emotion Estimation Method for Heart Rate Variability

2024-06-15T11:23:06+00:00

This study introduces a self-mental care management system utilizing an emotion estimation technique applied to heart rate variability parameters derived from vital data. The escalating prevalence of pandemics has exacerbated the dearth of accessible mental healthcare services, underscoring the heightened significance of investing in mental health programs. However, advancements in sensor technology, marked by enhanced precision and reduced size, facilitate the rapid acquisition of vital data from users. In our approach, we capture electrocardiogram (ECG) data concurrently with the narration of emotionally evocative stories. Through the analysis of the resulting data trends, emotions strongly correlated with the newly acquired ECG data were identified and inferred to be the emotions manifested within the ECG data. Our methodology enables the estimation of user emotions based on ECG data, and is further implemented in an application featuring real-time visualization of users' emotional states through chat icons. The deployment of this application empowers users to monitor emotional fluctuations and effectively manage their mental wellbeing.

Human Brain Judgment and Automated Classification of Masked Facial Expressions

2024-06-22T15:19:19+00:00

We investigated brain activity in response to facial expressions wearing masks. N170 responses at the T5 and T6 sites were synchronized with the vertex positive potential (VPP) response at the Cz site. The N170 responses were increased under masked face conditions, which may be associated with amodal completion. We then tested the facial emotion recognizer (FER) as a general classifier and the specifically created classifiers based on convolutional neural networks (CNNs) for predicting masked facial expressions. Although the accuracies in the FER were greatly lower for Japanese faces with masks than without masks, the specific CNN classifier improved the accuracies under the masked conditions.