Soonyoung Jung

• Korea University

Conditioned source separations have attracted significant attention because of their flexibility, applicability and extensionality. Their performance was usually inferior to the existing approaches, such as the single source separation model. However, a recently proposed method called LaSAFT-Net has shown that conditioned models can show comparable...

Recently, many methods based on deep learning have been proposed for music source separation. Some state-of-the-art methods have shown that stacking many layers with many skip connections improve the SDR performance. Although such a deep and complex architecture shows outstanding performance, it usually requires numerous computing resources and tim...

This paper proposes a neural network that performs audio transformations to user-specified sources (e.g., vocals) of a given audio track according to a given description while preserving other sources not mentioned in the description. Audio Manipulation on a Specific Source (AMSS) is challenging because a sound object (i.e., a waveform sample or fr...

Recent deep-learning approaches have shown that Frequency Transformation (FT) blocks can significantly improve spectrogram-based single-source separation models by capturing frequency patterns. The goal of this paper is to extend the FT block to fit the multi-source task. We propose the Latent Source Attentive Frequency Transformation (LaSAFT) bloc...

Singing Voice Separation (SVS) tries to separate singing voice from a given mixed musical signal. Recently, many U-Net-based models have been proposed for the SVS task, but there were no existing works that evaluate and compare various types of intermediate blocks that can be used in the U-Net architecture. In this paper, we introduce a variety of...

We consider an application scenario where user want to find regions that have similar tendency about a certain issue, e.g., looking for regions that are neutral to new welfare policies. Motivated by this, we present a novel query to retrieve regions with similar tendency, named ρ-Dense Region Query (ρ-DR Query), that returns arbitrary shape of regi...

Named entities in the biomedical domain are often written using a Noun Phrase (NP) along with a coordinating conjunction such as such as 'and' and 'or'. In addition, repeated words among named entity mentions are frequently omitted. It is often difficult to identify named entities. Although various Named Entity Recognition (NER) methods have tried...

Web search users complain of the inaccurate results produced by current search engines. Most of these inaccurate results are due to a failure to understand the user’s search goal. This paper proposes a method to extract users’ intentions and to build an intention map representing these extracted intentions. The proposed method makes intention vecto...

Digital content is easy to reproduce and manipulate. It is difficult to distinguish the original content from pirated copies. A cryptographic method is needed to protect content author’s ownership and secure content distribution. The method should be extended to the case of multiple authors since the content is completed with the assistance of many...

This paper proposes a novel of a personalized Computer Assisted Language Learning (CALL) system based on learner's cognitive abilities related to foreign language proficiency. In this CALL system, a strategy of retrieval learning, a method of learning memory cycle, and a method of repeated learning are applied for effective vocabulary memorization....

Web search users complain of inaccurate results of the current search engines. Most of inaccurate results are from failing to understand user's search goal. This paper proposes a method to mine user's intentions and to build an intention map representing their information needs. It selects intention features from search logs obtained from previous...

This paper proposes a computational lexical entry acquisition model based on a representation model of the mental lexicon. The proposed model acquires lexical entries from a raw corpus by unsupervised learning like human. The model is composed of full-form and morpheme acquisition modules. In the full-from acquisition module, core full-forms are au...

Hybrid Peer-to-Peer (P2P) systems that construct overlay networks structured among superpeers have great potential in that they can give the benefits such as scalability, search speed and network traffic, taking advantages of superpeer-based and the structured P2P systems. In this article, we enhance keyword search in hybrid P2P systems by construc...

This paper proposes a Computer Assisted Language Learning(CALL) system, the English Brain Enhancement System, in which is applied the brain scientific principles of human language processing for effective English learning. It is designed to offer adaptive learning based on the cognitive abilities related human language processing. For this, the sys...

As a solution for malicious Web page attacks, we need a client-side prevention system with deobfuscation mechanism. In this paper we introduce WebVaccine, that is a client-side real-time prevention system against obfuscated malicious Web pages. For secure web surfing, before the navigation to be completed, WebVaccine inserts a deobfuscation functio...

This study is intended to design and implement an automatic lexical acquisition model based on cognitive neuroscience referring to the theory that mental lexicon structure is represented with full-listing and morphemic and that lexical forms accessed to mental lexicon upon word cognition takes a hybrid type. As the result of the study, we could sim...

The HLA control a variety of function involved in immune response and influence susceptibility to over 40 diseases. It is important to find out how HLA cause the disease or modify susceptibility or course of it. In this paper, we developed an automatic HLA-disease information extraction procedure that uses biomedical publications. First, HLA and di...

The HLA control a variety of function involved in immune response and influence susceptibility to over 40 diseases. It is important to find out how HLA cause the disease or modify susceptibility or course of it. In this paper, we developed an automatic HLA-disease information extraction procedure that uses biomedical publications. First, HLA and di...

To facilitate power saving via wireless data broadcast, index information is typically broadcast along with the data. By first accessing the broadcast index, the mobile client is able to predict the arrival time of the desired data. However, it suffers from the drawback that the client has to wait and tune for an index segment, in order to conserve...

In this paper, we investigate whether the word frequency effect and the word similarity effect could be applied to Korean lexical decision task (henceforth, LDT). Also we propose a computational model of Korean LDT and present comparison results between human and computational model on Korean LDT. We found that the word frequency effect and the sim...

A peer-to-peer grid computing is complicated by heterogeneous capabilities, failures, volatility, and lack of trust because it is based on desktop computers at the edge of the Internet. In order to improve the reliability of computation and gain better performance, a replication mechanism must adapt to these distinct features. In other words, it is...

Peer-to-peer grid computing is an attractive computing paradigm for high throughput applications. However, both volatility Peer-to-peer grid computing is an attractive computing paradigm for high throughput applications. However, both volatility due to the autonomy of volunteers (i.e., resource providers) and the heterogeneous properties of volunte...

In this paper, we investigate whether the word frequency effect and the word similarity effect could be applied to Korean lexical decision task (henceforth, LDT). Also we propose a computational model of Korean LDT and present comparison results between human and computational model on Korean LDT. We found that the word frequency effect and the sim...

Location-aware mobile services(LAMSs) are characterized by a large number of objects and a large number of queries. Moreover, with a large candidate data set, answering LAMSs via scanning through the whole data set becomes extremely expensive. In broadcast-based services, any number of clients can monitor the broadcast channel and retrieve the data...

In a peer to peer grid computing environment, volunteers have heterogeneous properties and dynamically join and leave during execution. Therefore, it is essential to adapt to an unstable and widely distributed environment. However, existing scheduling and result certification mechanisms do not adapt to such a dynamic environment. As a result, they...

Indexing techniques have been developed for wireless data broadcast environments, in order to conserve the scarce power resources of the mobile clients. However, the use of interleaved index segments in a broadcast cycle increases the average access latency for the clients. In this paper, we present the broadcast-based spatial query processing algo...

The paper proposes a new conflict relation called preservation. By explicitly including reverse-operations, preservation can be used independently of execution contexts to which different recovery algorithms and/or object models give rise. The paper also makes a two-dimensional (i.e., execution contexts and operations' specifications) comparison be...