• Journal of Korea Multimedia Society
• /
• v.22 no.9
• /
• pp.1000-1010
• /
• 2019

Image semantic segmentation and dehazing are key tasks in the computer vision. In recent years, researches in both tasks have achieved substantial improvements in performance with the development of Convolutional Neural Network (CNN). However, most of the previous works for semantic segmentation assume the images are captured in clear weather and show degraded performance under hazy images with low contrast and faded color. Meanwhile, dehazing aims to recover clear image given observed hazy image, which is an ill-posed problem and can be alleviated with additional information about the image. In this work, we propose a deep multi-task network for simultaneous semantic segmentation and dehazing. The proposed network takes single haze image as input and predicts dense semantic segmentation map and clear image. The visual information getting refined during the dehazing process can help the recognition task of semantic segmentation. On the other hand, semantic features obtained during the semantic segmentation process can provide cues for color priors for objects, which can help dehazing process. Experimental results demonstrate the effectiveness of the proposed multi-task approach, showing improved performance compared to the separate networks.

• The Journal of the Korea Contents Association
• /
• v.9 no.3
• /
• pp.1-8
• /
• 2009

This paper proposes a new document clustering system using fuzzy logic-based genetic algorithm (GA) and semantic vector expansion technology. It has been known in many GA papers that the success depends on two factors, the diversity of the population and the capability to convergence. We use the fuzzy logic-based operators to adaptively adjust the influence between these two factors. In traditional document clustering, the most popular and straightforward approach to represent the document is vector space model (VSM). However, this approach not only leads to a high dimensional feature space, but also ignores the semantic relationships between some important words, which would affect the accuracy of clustering. In this paper we use latent semantic analysis (LSA)to expand the documents to corresponding semantic vectors conceptually, rather than the individual terms. Meanwhile, the sizes of the vectors can be reduced drastically. We test our clustering algorithm on 20 news groups and Reuter collection data sets. The results show that our method outperforms the conventional GA in various document representation environments.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 1996.02a
• /
• pp.157-166
• /
• 1996

This paper treats the japanese adjective phrase forming derivational suffix -tai from a new point of view: firstly it tries to approach from a semantic standpoint by applying the proposal made in Ikeya (1991). It will be shown that adjective phrases formed by -tai fits nicely with the semantic structure proposed by Ikeya. Secondly, we attempt to 'derive' -tai sentences by adopting a basic framework of HPSG so that we can 'derive' them without having recourse to transformational operations, that is, in a monostratal way. In tackling the problem we have tried to incorporate many ideas proposed so far on this issue.

• Language and Information
• /
• v.17 no.2
• /
• pp.29-50
• /
• 2013

Multiple Nominative Constructions (MNCs) and Multiple Accusative Constructions (MACs) have been some of the hottest and interesting topics in Korean syntax. This paper took empirical approaches to these constructions and examined native speakers' grammaticality judgements of these constructions. Though there are lots of previous studies on these constructions, Ryu (2010, 2013a, 2013b, 2013c) recently tried to unify MNCs and MACs into Multiple Case Constructions (MCCs) and to classify them into 16 types based on the semantic relations. This paper includes experiments which were performed on these 16 different types. The experiments were designed following Johnson (2008); and the native speakers' intuition was measured with two scales, numerical estimates and line drawing, though the latter was adopted in the actual analyses. Through the experiment, the following facts were observed: (i) the grammaticality of the MCCs varies depending on their semantic relations, (ii) MNCs were more grammatical than MACs if both constructions occurred in similar environments, and (iii) the sentences in some MAC types had much lower grammaticality than those in the others, as Ryu (2013b, 2013c) mentioned.

• Proceedings of the KSPS conference
• /
• 2003.10a
• /
• pp.153-156
• /
• 2003

Voice input is often required in many new application environments such as telephone-based information retrieval, car navigation systems, and user-friendly interfaces, but the low success rate of speech recognition makes it difficult to extend its application to new fields. Popular approaches to increase the accuracy of the recognition rate have been researched by post-processing of the recognition results, but previous approaches were mainly lexical-oriented ones in post error correction. We suggest a new semantic-oriented approach to correct both semantic level and lexical errors, which is also more accurate for especially domain-specific speech error correction. Through extensive experiments using a speech-driven in-vehicle telematics information application, we demonstrate the superior performance of our approach and some advantages over previous lexical-oriented approaches.

• Language and Information
• /
• v.4 no.1
• /
• pp.21-42
• /
• 2000

In this paper, I explore the semantic interpretation of the English present perfect by arguing that the perfect is analogous to modals in its interpretation. The perfect produces several different readings, i.e., the resultative and the current relevant reading, to mention a few. Despite this, the meaning of the perfect remains invariable in sentences where it occurs. Instead, the semantic variability of the perfect is due to the nature of the conversational background. This indicates that just as modals are context-dependent, so is the perfect, which inspires a modal-based approach to the semantics of the perfect. By incorporating such an approach into its semantic analysis, we can present a unified account of the different meanings of the perfect.

• English Language & Literature Teaching
• /
• no.3
• /
• pp.185-201
• /
• 1997

This study explores perspectives on responding to EFL students' compositions at the semantic level. In the last three decades, there has been a shift from product-oriented approach to process-oriented one to teaching writing. The shift has led to the criticism of the traditional view on teacher response. The traditional view has been under attack for its overemphasis upon form and ineffectiveness on improving student writing skill. It is also noted that research into students' reactions to the traditional teacher response has been inconclusive. The process-oriented approach, on the other hand, draws its attention to meaning and the logical development of thought as well as linguistic matters. In this context, the present study discusses what EFL teachers need to take into account in providing the semantic-level feedback on students' compositions. Firstly, teacher response to student writing is on-going; teacher feedback involves teacher intervention in the drafting process, the revision process, and the presentation of product. Secondly, in the writing conferences, the teacher provides students an opportunity to talk about writing, assistance and advice on the content/meaning of the written text, helping them expand and clarify thinking about audience(reader) and purpose.

• ETRI Journal
• /
• v.34 no.2
• /
• pp.288-291
• /
• 2012

In this letter, a new online anomaly detection approach for software systems is proposed. The novelty of the proposed approach is to apply a new semantic kernel function for a support vector machine (SVM) classifier to detect fault-suspicious execution paths at runtime in a reasonable amount of time. The kernel uses a new sequence matching algorithm to measure similarities among program execution paths in a customized feature space whose dimensions represent the largest common subpaths among the execution paths. To increase the precision of the SVM classifier, each common subpath is given weights according to its ability to discern executions as correct or anomalous. Experiment results show that compared with the known kernels, the proposed SVM kernel will improve the time overhead of online anomaly detection by up to 170%, while improving the precision of anomaly alerts by up to 140%.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.13 no.2
• /
• pp.102-110
• /
• 2017

Current word sense disambiguation techniques employ various machine learning-based methods. Various approaches have been proposed to address this problem, including the knowledge base approach. This approach defines the sense of an ambiguous word in accordance with knowledge base information with no training corpus. In unsupervised learning techniques that use a knowledge base approach, graph-based and similarity-based methods have been the main research areas. The graph-based method has the advantage of constructing a semantic graph that delineates all paths between different senses that an ambiguous word may have. However, unnecessary semantic paths may be introduced, thereby increasing the risk of errors. To solve this problem and construct a fine-grained graph, in this paper, we propose a model that iteratively constructs the graph while eliminating unnecessary nodes and edges, i.e., senses and semantic paths. The hybrid similarity estimation model was applied to estimate a more accurate sense in the constructed semantic graph. Because the proposed model uses BabelNet, a multilingual lexical knowledge base, the model is not limited to a specific language.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.12
• /
• pp.5782-5799
• /
• 2018

With the development of GPS and the popularity of mobile devices with positioning capability, collecting massive amounts of trajectory data is feasible and easy. The daily trajectories of moving objects convey a concise overview of their behaviors. Different social roles have different trajectory patterns. Therefore, we can identify users or groups based on similar trajectory patterns by mining implicit life patterns. However, most existing daily trajectories mining studies mainly focus on the spatial and temporal analysis of raw trajectory data but missing the essential semantic information or behaviors. In this paper, we propose a novel trajectory semantics calculation method to identify groups that have similar behaviors. In our model, we first propose a fast and efficient approach for stay regions extraction from daily trajectories, then generate semantic trajectories by enriching the stay regions with semantic labels. To measure the similarity between semantic trajectories, we design a semantic similarity measure model based on spatial and temporal similarity factor. Furthermore, a pruning strategy is proposed to lighten tedious calculations and comparisons. We have conducted extensive experiments on real trajectory dataset of Geolife project, and the experimental results show our proposed method is both effective and efficient.