• Proceedings of the KSPS conference
• /
• 2005.04a
• /
• pp.49-52
• /
• 2005

Cepstral Mean Subtraction (CMS) makes effectively compensation for a channel distortion, but there are some shortcomings such as distortions of feature parameters, waiting for the whole speech sentence. By assuming that the silence parts have the channel characteristics, we consider the channel normalization using subtraction of cepstral means which are only obtained in the silence areas. If the considered techniques are successfully used for the channel compensation, the proposed method can be used for real time processing environments or time important areas. In the experiment result, however, the performance of our method is not good as CMS technique. From the analysis of the results, we found potentiality of the proposed method and will try to find the technique reducing the gap between CMS and ours method.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 2007.11a
• /
• pp.385-393
• /
• 2007

Corpora annotated with lots of linguistic information are required to develop robust and statistical natural language processing systems. Building such corpora, however, is an expensive, labor-intensive, and time-consuming work. To help the work, we design and implement an annotation tool for establishing a Korean dependency tree-tagged corpus. Compared with other annotation tools, our tool is characterized by the following features: independence of applications, localization of errors, powerful error checking, instant annotated information sharing, user-friendly. Using our tool, we have annotated 100,904 Korean sentences with dependency structures. The number of annotators is 33, the average annotation time is about 4 minutes per sentence, and the total period of the annotation is 5 months. We are confident that we can have accurate and consistent annotations as well as reduced labor and time.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.4E
• /
• pp.26-30
• /
• 1999

This paper introduces a prosodic phrasing method in Korean to improve the naturalness of speech synthesis, especially in text-to-speech conversion. In prosodic phrasing, it is necessary to understand the structure of a sentence through a language processing procedure, such as part-of-speech (POS) tagging and parsing, since syntactic structure correlates better with the prosodic structure of speech than with other factors. In this paper, the prosodic phrasing procedure is treated from two perspectives: dependency parsing and prosodic phrasing using dependency relations. This is appropriate for Ural-Altaic, since a prosodic boundary in speech usually concurs with a governor of dependency relation. From experimental results, using the proposed method achieved 12% improvement in prosody boundary prediction accuracy with a speech corpus consisting 300 sentences uttered by 3 speakers.

• International Journal of Contents
• /
• v.5 no.3
• /
• pp.65-70
• /
• 2009

This paper describes ongoing work on the correction of adverb errors committed by Korean learners studying English as a foreign language (EFL), using an automated English writing assessment system. Adverb errors are commonly found in learners 'writings, but handling those errors rarely draws an attention in natural language processing due to complicated characteristics of adverb. To correctly detect the errors, adverbs are classified according to their grammatical functions, meanings and positions within a sentence. Adverb errors are collected from learners' sentences, and classified into five categories adopting a traditional error analysis. The error classification in conjunction with the adverb categorization is implemented into a set of mal-rules which automatically identifies the errors. When an error is detected, the system corrects the error and suggests error specific feedback. The feedback includes the types of errors, a corrected string of the error and a brief description of the error. This attempt suggests how to improve adverb error correction method as well as to provide richer diagnostic feedback to the learners.

• Proceedings of the KIEE Conference
• /
• 2003.11c
• /
• pp.836-839
• /
• 2003

In this paper, we implement a language model by a bigram and evaluate proper smoothing technique for unit of low perplexity. Word, morpheme, clause units are widely used as a language processing unit of the language model. We propose VCCV units which have more small vocabulary than morpheme and clauses units. We compare the VCCV units with the clause and the morpheme units using the perplexity. The most common metric for evaluating a language model is the probability that the model assigns the derivative measures of perplexity. Smoothing used to estimate probabilities when there are insufficient data to estimate probabilities accurately. In this paper, we constructed the N-grams of the VCCV units with low perplexity and tested the language model using Katz, Witten-Bell, absolute, modified Kneser-Ney smoothing and so on. In the experiment results, the modified Kneser-Ney smoothing is tested proper smoothing technique for VCCV units.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.8 no.3
• /
• pp.225-230
• /
• 2008

In this paper we embody a system that recognizes printed alphabet, numeral figures and symbols written on the keyboard for the recognition of English sentences. The image of the printed sentences is inputted and binarized, and the characters are separated by using histogram method that is the same as the existing character recognition method. During the abstraction of the individual characters, we classify one group that has not numerical information by the projection of the vertical center of the character. In case of another group that has the longer width than the height, we assort them by normalizing the width. The other group normalizes the height of the images. With the reverse application of the basic principle of the Japanese Puzzle to a normalized character image, the proposed system classifies and recognizes the printed numeral figures, symbols and characters, consequently we meet with good result.

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.1
• /
• pp.17-24
• /
• 2018

With improvement of speech recognition and natural language processing, dialog systems are recently adapted to various service domains. It became possible to get desirable services by conversation through the dialog system, but it is still necessary to improve separate modules, such as domain detection, intention detection, named entity recognition, and out-of-domain detection, in order to achieve stable service offer. When it misclassifies an in-domain sentence of conversation as out-of-domain, it will result in poor customer satisfaction and finally lost business. As there have been relatively small number of studies related to the out-of-domain detection, in this paper, we introduce a new method using a hierarchical Dirichlet process and demonstrate the effectiveness of it by experimental results on Korean dataset.

• Journal of Korea Multimedia Society
• /
• v.24 no.5
• /
• pp.651-658
• /
• 2021

In this paper, we propose methods for character recognition and classification for table items through deep learning. First, table areas are detected in a document image through CNN. After that, table areas are separated by separators such as vertical lines. The text in document is recognized through a neural network combined with CNN and RNN. To correct errors in the character recognition, multiple candidates for the recognized result are provided for a sentence which has low recognition accuracy.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.1
• /
• pp.81-88
• /
• 2021

Recently, many videos such as online lecture videos caused by COVID-19 have been generated. However, due to the limitation of working hours and lack of cost, they are only a part of the videos with subtitles. It is emerging as an obstructive factor in the acquisition of information by deaf. In this paper, we try to develop a system that automatically generates subtitles using voice recognition and generates subtitles by separating sentences using the ending and time to reduce the time and labor required for subtitle generation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.305-318
• /
• 2022

Abusive behaviors have become a common issue in many online social media platforms. Profanity is common form of abusive behavior in online. Social media platforms operate the filtering system using popular profanity words lists, but this method has drawbacks that it can be bypassed using an altered form and it can detect normal sentences as profanity. Especially in Korean language, the syllable is composed of graphemes and words are composed of multiple syllables, it can be decomposed into graphemes without impairing the transmission of meaning, and the form of a profane word can be seen as a different meaning in a sentence. This work focuses on the problem of filtering system mis-detecting normal phrases with profane phrases. For that, we proposed the deep learning-based framework including grapheme and syllable separation-based word embedding and appropriate CNN structure. The proposed model was evaluated on the chatting contents from the one of the famous online games in South Korea and generated 90.4% accuracy.