Chae-Gyun Lim

Analyzing emotions in twitter during a crisis: A case study of the 2015 Middle East Respiratory Syndrome outbreak in Korea

Social media has emerged as an effective source to investigate peoples opinions in the context of a variety of topics and situations. In particular, many recent studies try to investigate social media during the crisis situations that range from natural disasters to man-made conflicts. In this paper, we investigate peoples emotional responses expressed on Twitter during the 2015 Middle East Respiratory Syndrome (MERS) outbreak in South Korea. Specifically, we first present an emotion analysis method to classify fine-grained emotions in Korean Twitter posts. Then, we conduct a case study of how Korean Twitter users responded to MERS outbreak using our emotion analysis method. Experimental results on Korean benchmark dataset demonstrate the superior performance of the proposed emotion analysis approach on real-world dataset. Moreover, our analysis results on tweets related to MERS outbreak help to understand the behaviors of humans and the characteristics of sociocultural system. Further, our method can be harnessed by the media to automatically investigate public opinions as well as the authorities to gain insights for quickly deciding the assistance policies.

Temporal Information Extraction from Korean Texts

As documents tend to contain temporal information, extracting such information is attracting much research interests recently. In this paper, we propose a hybrid method that combines machine-learning models and hand-crafted rules for the task of extracting temporal information from unstructured Korean texts. We address Korean-specific research issues and propose a new probabilistic model to generate complementary features. The performance of our approach is demonstrated by experiments on the TempEval-2 dataset, and the Korean TimeBank dataset which we built for this study.

A survey of temporal information extraction and language independent features

Since there has been an explosive growth of online documents, the knowledge extraction from natural language texts becomes more important to complement limited human ability. In this paper, we introduce existing methods for temporal information extraction from input texts in two viewpoints. One is the researches about language-independent feature generation. Even though linguistic features have been generally utilized to extract time expressions, we need different features which are independent from language characteristics in order to overcome the boundary of particular language. Another is the researches about temporal information extraction based on the common patterns or knowledge bases. By summarizing and discussing existing researches, we can see the current research direction on this field to help better understanding of the temporal information extraction methods.

Modeling long-term human activeness using recurrent neural networks for biometric data

BackgroundWith the invention of fitness trackers, it has been possible to continuously monitor a user’s biometric data such as heart rates, number of footsteps taken, and amount of calories burned. This paper names the time series of these three types of biometric data, the user’s “activeness”, and investigates the feasibility in modeling and predicting the long-term activeness of the user.MethodsThe dataset used in this study consisted of several months of biometric time-series data gathered by seven users independently. Four recurrent neural network (RNN) architectures–as well as a deep neural network and a simple regression model–were proposed to investigate the performance on predicting the activeness of the user under various length-related hyper-parameter settings. In addition, the learned model was tested to predict the time period when the user’s activeness falls below a certain threshold.ResultsA preliminary experimental result shows that each type of activeness data exhibited a short-term autocorrelation; and among the three types of data, the consumed calories and the number of footsteps were positively correlated, while the heart rate data showed almost no correlation with neither of them. It is probably due to this characteristic of the dataset that although the RNN models produced the best results on modeling the user’s activeness, the difference was marginal; and other baseline models, especially the linear regression model, performed quite admirably as well. Further experimental results show that it is feasible to predict a user’s future activeness with precision, for example, a trained RNN model could predict–with the precision of 84%–when the user would be less active within the next hour given the latest 15 min of his activeness data.ConclusionsThis paper defines and investigates the notion of a user’s “activeness”, and shows that forecasting the long-term activeness of the user is indeed possible. Such information can be utilized by a health-related application to proactively recommend suitable events or services to the user.

An extension of the automatic cross-association method with a 3-dimensional matrix

There are numerous 2-dimensional matrix data for clustering including a set of documents, citation networks, web graphs, etc. However, many real-world datasets have more than three modes which require at least 3-dimensional matrices or tensors. Focusing on the clustering algorithm known as cross-association, we extend the algorithm to deal with a 3-dimensional matrix. Our proposed method is fully automated, and simultaneously discovers clusters of both row, column, and tube groups. Experiments on real and synthetic datasets show that our method is effective. Through the proposed method, useful information can be obtained even from sparse datasets.

Context-based healthy lifestyle recommendation for enhancing user's wellness

In this paper, we propose a service to predict changes in a users daily activity in consideration of the users contextual information and recommend appropriate physical activities for improving well-being. Based on the data collected from a smartphone and wearable sensor, our service models an activity pattern of the specific user and predicts the upcoming activity changes of the user by using the individuals lifestyle model. In addition, we continually observe changes in daily activity to determine appropriate preferred activities for each user, and recommend optimal activities to promote wellness by tracking changes in contextual information and user feedback. It can be confirmed that the use of the proposed service contributes to the improvement of the users wellness by performing a clinical experiment on the actual subjects.

A center-biased graph learning algorithm for image classification

Graph matching is an important problem in the field of computer vision. Graph matching problem can be represented as quadratic assignment problem. Because the problem is known to be NP-hard, optimal solution is hardly achievable so that a lot of algorithms are proposed to approximate it. Although there have been many studies about fast and accurate approximations, there have been few studies about graph learning. This paper presents a graph learning algorithm which works in an unsupervised way. The process requires neither annotated dataset nor training dataset. The algorithm learns a graph from a model image using a variation of random walk, which we call center biased random walk with restart (CBRWR). This algorithm can be implemented using two-dimensional Gaussian distribution. For this, we propose a modified histogram-based attribute. The attributes consider relationship between edges as well as nodes. Image matching is done using the model graph which is created by our method. We conducted image classification experiments to check the competitiveness of our algorithm.

Investigating the Impact of Possession-Way of a Smartphone on Action Recognition

For the past few decades, action recognition has been attracting many researchers due to its wide use in a variety of applications. Especially with the increasing number of smartphone users, many studies have been conducted using sensors within a smartphone. However, a lot of these studies assume that the users carry the device in specific ways such as by hand, in a pocket, in a bag, etc. This paper investigates the impact of providing an action recognition system with the information of the possession-way of a smartphone, and vice versa. The experimental dataset consists of five possession-ways (hand, backpack, upper-pocket, lower-pocket, and shoulder-bag) and two actions (walking and running) gathered by seven users separately. Various machine learning models including recurrent neural network architectures are employed to explore the relationship between the action recognition and the possession-way recognition. The experimental results show that the assumption of possession-ways of smartphones do affect the performance of action recognition, and vice versa. The results also reveal that a good performance is achieved when both actions and possession-ways are recognized simultaneously.