Yu-Seop Kim

An empirical study on dimensionality optimization in text mining for linguistic knowledge acquisition

In this paper, we try to find empirically the optimal dimensionality in data-driven models, Latent Semantic Analysis (LSA) model and Probabilistic Latent Semantic Analysis (PLSA) model. These models are used for building linguistic semantic knowledge which could be used in estimating contextual semantic similarity for the target word selection in English-Korean machine translation. We also facilitate k-Nearest Neighbor learning algorithm. We diversify our experiments by analyzing the covariance between the value of k in k-NN learning and accuracy of selection, in addition to that between the dimensionality and the accuracy. While we could not find regular tendency of relationship between the dimensionality and the accuracy, however, we could find the optimal dimensionality having the most sound distribution of data during experiments.

A First Person Shooter with Dual Guns Using Multiple Optical Air Mouse Devices

We proposed a first person shooter for both hands with multiple mouse devices, which are not used in existing games. Players control their movement and shooting separately using dual guns for both hands. For two-hand usage, we designed a display method for multiple mouse pointers. First, we loaded typical physical driver into the memory and registered our developed multiple mouse device driver instead of the existing MS Windows device driver. We then executed both message loop procedures for Windows mouse event handling and those for multiple mouse event handling. We could display multiple mouse cursor images and utilize both mouse devices separately. After that, a prototype game using both hands was developed for an experimental study.

Mandibular second molar root canal morphology and variants in a Korean subpopulation.

AIM To determine the root canal anatomy of mandibular second molars in a Korean population by analysing cone-beam computed tomography (CBCT) images. METHODOLOGY The CBCT images of 960 subjects were examined. The number and configuration of roots and canals were categorized according to Vertuccis and modified Meltons classifications. RESULTS Of the 1920 mandibular second molars, 41% had one root, 58% had two roots, and <1% had three roots. In the mesial roots of two-rooted molars, Vertuccis Type lV (44%) and Type II (37.75%) canals were most frequent. The prevalence of C-shaped roots was 40%, and C-shaped roots in combination with additional mesiolingual or distolingual roots were found in <1% of molars. Interestingly, O-shaped canals were detected in 0.10% of the molars. Of the C-shaped roots, the most common configuration types were Meltons Type I (66%) in the coronal region and Meltons Type III (56%) in the apical region. The prevalence of C-shaped roots was higher in females (47%) than in males (32%) (P < 0.001) and did not differ with age (P = 0.497) or tooth position (P = 0.514). Most (82%) C-shaped canals were bilateral (P < 0.001). CONCLUSIONS A high prevalence of C-shaped canals and a low incidence of three-rooted molars were observed in the mandibular second molars of the Korean subpopulation studied.

Training a Korean SRL System with Rich Morphological Features

In this paper we introduce a semantic role labeler for Korean, an agglutinative language with rich morphology. First, we create a novel training source by semantically annotating a Korean corpus containing fine-grained morphological and syntactic information. We then develop a supervised SRL model by leveraging morphological features of Korean that tend to correspond with semantic roles. Our model also employs a variety of latent morpheme representations induced from a larger body of unannotated Korean text. These elements lead to state-of-the-art performance of 81.07% labeled F1, representing the best SRL performance reported to date for an agglutinative language.

A comparative evaluation of data-driven models in translation selection of machine translation

We present a comparative evaluation of two data-driven models used in translation selection of English-Korean machine translation. Latent semantic analysis(LSA) and probabilistic latent semantic analysis (PLSA) are applied for the purpose of implementation of data-driven models in particular. These models are able to represent complex semantic structures of given contexts, like text passages. Grammatical relationships, stored in dictionaries, are utilized in translation selection essentially. We have used k-nearest neighbor (k-NN) learning to select an appropriate translation of the unseen instances in the dictionary. The distance of instances in k-NN is computed by estimating the similarity measured by LSA and PLSA. For experiments, we used TREC data(AP news in 1988) for constructing latent semantic spaces of two models and Wall Street Journal corpus for evaluating the translation accuracy in each model. PLSA selected relatively more accurate translations than LSA in the experiment, irrespective of the value of k and the types of grammatical relationship.

Topic Extraction from Text Documents Using Multiple-Cause Networks

This paper presents an approach to the topic extraction from text documents using probabilistic graphical models. Multiple-cause networks with latent variables are used and the Helmholtz machines are utilized to ease the learning and inference. The learning in this model is conducted in a purely data-driven way and does not require prespecified categories of the given documents. Topic words extraction experiments on the TDT-2collection are presented. Especially, document clustering results on a subset of TREC-8 ad-hoc task data show the substantial reduction of the inference time without significant deterioration of performance.

An intelligent grading system using heterogeneous linguistic resources

In this paper, we propose an intelligent grading system using heterogeneous linguistic resources. We used latent semantic kernel as one resource in former research and found that a deficit of indexed terms gave rise to performance bottleneck. To solve this, we expand answer papers, written by students and instructors, by utilizing one of widely used linguistic resources, WordNet. We supplement the papers with words semantically related to indexed terms of papers. The added words are selected from the synonyms and hyponyms on WordNet. And to get rid of the criterion decision problem, we use partial score of each question and evaluate the correlation coefficient between grading results of the proposed approach and human instructors. The proposed approach in this research achieves maximally 0.94 correlation coefficient to instructors, which is 0.06 higher than that of the former research.

Micropatterning of bacteria on two-dimensional lattice protein surface observed by atomic force microscopy.

In this study, we characterized the two-dimensional lattice of bovine serum albumin (BSA) as a chemical and physical barrier against bacterial adhesion, using fluorescence microscopy and atomic force microscopy (AFM). The lattice of BSA on glass surface was fabricated by micro-contact printing (microCP), which is a useful way to pattern a wide range of molecules into microscale features on different types of substrates. The contact-mode AFM measurements showed that the average height of the printed BSA monolayer was 5-6 nm. Escherichia coli adhered rapidly on bare glass slide, while the bacterial adhesion was minimized on the lattices in the range of 1-3 microm(2). Especially, the bacterial adhesion was completely inhibited on a 1 microm(2) lattice. The results suggest that the anti-adhesion effects are due by the steric repulsion forces exerted by BSA.

Intra-sentence segmentation based on support vector machines in English-Korean machine translation systems

This work is about intra-sentence segmentation performed before syntactic analysis of long sentences composed of at least 20 words in an English-Korean machine translation system. A long sentence has been known to spend enormous computational time and space when it is analyzed syntactically. It can also produce poor translation results. To resolve this problem, we partitioned a long sentence into a few segments to analyze each segment separately. To partition the sentence, firstly, we tried to find candidates for each segment position in the sentence. We then generated input vectors representing lexical contexts of the corresponding candidates and also used the support vector machines (SVM) algorithm to learn and recognize the appropriate segment positions. We used three kernel functions, the linear kernel, the polynomial kernel and the Gaussian kernel, to find optimal hyperplanes classifying proper positions and we compared results obtained from each kernel function. As a result of the experiments, we acquired 0.81, 0.83, and 0.79 f-measure values from the linear, polynomial and Gaussian kernel, respectively.

An autonomous assessment system based on combined latent semantic kernels

In this paper, we develop an autonomous assessment system based on the kernel combinations which are mixed by two kernel matrices from the WordNet and corpus. Many researchers have tried to integrate these two resources in many applications, to utilize diverse information extracted from each resource. However, since two resources have been represented in quite different ways, one resource has been secondary to another. To fully integrate two resources at the same level, we first transform the WordNet, which has a hierarchical structure, into a matrix structure. Concurrently, another matrix, which represents a co-occurrence of words in the collection of text documents, is constructed. We then build two initial latent semantic kernels from both matrices and merge them into a new single kernel matrix. When we merge two matrices, we split each initial matrix into independent columns and mix the columns with various methods. We acquire a few combined kernel matrices which show various performances in experiments. Compared to the basic vector space model, original kernel matrices, and the BLEU based method, the combined matrices improve the accuracy of assessment.