• The Journal of the Acoustical Society of Korea
• /
• v.14 no.4
• /
• pp.46-63
• /
• 1995

This thesis described an unlimited vocabulary connected speech recognition system using Time Delay Neural Network(TDNN). The recognition unit is the diphone unit which includes the transition section of two phonemes, and the number of diphone unit is 329. The recognition processing of korean connected speech is composed by three part; the feature extraction section of the input speech signal, the diphone recognition processing and post-processing. In the feature extraction section, the extraction of diphone interval in input speech signal is carried and then the feature vectors of 16th filter-bank coefficients are calculated for each frame in the diphone interval. The diphone recognition processing is comprised by the three stage hierachical structure and is carried using 30 Time Delay Neural Networks. particularly, the structure of TDNN is changed so as to increase the recognition rate. The post-processing section, mis-recognized diphone strings are corrected using the probability of phoneme transition and the probability o phoneme confusion and then the eojeols (Korean word or phrase) are formed by combining the recognized diphones.

• The Journal of the Korea Contents Association
• /
• v.17 no.4
• /
• pp.576-584
• /
• 2017

Unknown nouns which are not in a dictionary make problems not only morphological analysis but also almost all natural language processing area. This paper describes a recognition method for Korean unknown nouns using strings following nouns such as postposition, suffix and postposition, suffix and eomi, etc. We collect and sort words including nouns from documents and divide a word including unknown noun into two parts, candidate noun and string following the noun, by finding same prefix morphemes from more than two unknown words. We use information of strings following nouns extracted from Sejong corpus and decide unknown noun finally. We obtain 99.64% precision and 99.46% recall for unknown nouns occurred more than two forms in news of two portal sites.

• Korean Journal of Cognitive Science
• /
• v.18 no.4
• /
• pp.325-350
• /
• 2007

This study investigated global and local characteristics of eye movement while 16 college students read 48 easy Korean sentences. It was found that readers lusted for about 225ms at the word cluster(eojeol), made a forward saccade of about 3.6 characters to the next word, skipped short and high-frequent words about 25% during the first-pass reading, and regressed backward at 19%. There were also individual differences in readers' pattern of fixation and saccade. In addition, the effects of word cluster length and word frequency and the effects related to landing position were examined. The eyes landed on the center of a word cluster more frequently than on the boundaries. When the eyes landed at the boundaries, the eyes fixated the word cluster again more frequently. The word clusters with high-frequency words were read faster than those with low-frequency words.

• Journal of KIISE:Software and Applications
• /
• v.30 no.3_4
• /
• pp.358-371
• /
• 2003

Automatic word spacing is a process of deciding correct boundaries between words in a sentence including spacing errors. It is very important to increase the readability and to communicate the accurate meaning of text to the reader. The previous statistical approaches for automatic word spacing do not consider the previous spacing state, and thus can not help estimating inaccurate probabilities. In this paper, we propose two statistical word spacing models which can solve the problem of the previous statistical approaches. The proposed models are based on the observation that the automatic word spacing is regarded as a classification problem such as the POS tagging. The models can consider broader context and estimate more accurate probabilities by generalizing hidden Markov models. We have experimented the proposed models under a wide range of experimental conditions in order to compare them with the current state of the art, and also provided detailed error analysis of our models. The experimental results show that the proposed models have a syllable-unit accuracy of 98.33% and Eojeol-unit precision of 93.06% by the evaluation method considering compound nouns.

• Journal of KIISE:Software and Applications
• /
• v.35 no.11
• /
• pp.681-691
• /
• 2008

In this paper, we have evaluated and compared each feature and feature combinations necessary for statistical Korean dialogue act classification. We have implemented a Korean dialogue act classification system by using the Support Vector Machine method. The experimental results show that the POS bigram does not work well and the morpheme-POS pair and other features can be complementary to each other. In addition, a small number of features, which are selected by a feature selection technique such as chi-square, are enough to show steady performance of dialogue act classification. We also found that the last eojeol plays an important role in classifying an entire sentence, and that Korean characteristics such as free order and frequent subject ellipsis can affect the performance of dialogue act classification.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.10
• /
• pp.25-33
• /
• 2012

In this paper, we propose a rule-based model to correct errors in a speech recognition result in the mobile device environment. The proposed model considers the mobile device environment with limited resources such as processing time and memory, as follows. In order to minimize the error correction processing time, the proposed model removes some processing steps such as morphological analysis and the composition and decomposition of syllable. Also, the proposed model utilizes the longest match rule selection method to generate one error correction candidate per point, assumed that an error occurs. For the purpose of deploying memory resource, the proposed model uses neither the Eojeol dictionary nor the morphological analyzer, and stores a combined rule list without any classification. Considering the modification and maintenance of the proposed model, the error correction rules are automatically extracted from a training corpus. Experimental results show that the proposed model improves 5.27% on the precision and 5.60% on the recall based on Eojoel unit for the speech recognition result.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.366-377
• /
• 2005

We find a module may that takes optimal correction rate of Korean spelling corrector. If there are a lot of module numbers of spelling corrector, it is difficult to calculate optimal correction rate of spelling corrector because permutation of N-modules is N!. This Korean spelling corrector consists of 19 modules. It is impossible to arrange 19 modules actually and the correction rate is various according to input data. We found the range of correction rate using parallel processing between modules and the optimal correction rate using sequential processing of modules. Input data that are used in an experiment is 753,191 eojeol's sets that happen in newspaper publishing company during several years. About this error set, theoretical maximum correction rate of spelling corrector is $97.28\%$ (732,764/753,191). But we got the optimal correction rate $96.62\%$ (727,750/733,191). This optimal correction rate is almost near to $99.31\%$ (727,750/732,764) of the maximum correction rate.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.16 no.8
• /
• pp.5565-5570
• /
• 2015

Implicit citation sentence recognition is to locate citation sentences which lacks explicit citation markers, from articles' full-text. State-of-the-art approaches exploit word ngrams, clue words, researcher's surnames, mentions of previous methods, and distance relative to nearest explicit citation sentences, etc., reaching over 50% performance. However, most previous works have been conducted on English. As for Korean, a rule-based method using positive/negative clue patterns was reported to attain the performance of 42%, requiring further improvement. This study attempted to learn to recognize implicit citation sentences from Korean literatures' full-text using Korean lexical features. Different lexical feature units such as Eojeol, morpheme, and Eumjeol were evaluated to determine proper lexical features for Korean implicit citation sentence recognition. In addition, lexical features were combined with the position features representing backward/forward proximities to explicit citation sentences, improving the performance up to over 50%.

• The KIPS Transactions:PartB
• /
• v.10B no.1
• /
• pp.47-56
• /
• 2003

Korean delimits words by white-space like English, but words In Korean Is a little different in structure from those in English. Words in English generally consist of one word, but those in Korean are composed of one word and/or morpheme or more. Because of this difference, a word between white-spaces is called an Eojeol in Korean. We propose a method for segmenting and classifying Korean words and/or morphemes based on syllables using an instance-based learning. In this paper, elements of feature sets for the instance-based learning are one previous syllable, one current syllable, two next syllables, a final consonant of the current syllable, and two previous categories. Our method shows more than 97% of the F-measure of word segmentation using ETRI corpus and KAIST corpus.

• The Journal of the Korea Contents Association
• /
• v.22 no.3
• /
• pp.134-140
• /
• 2022

The purpose of this study was to investigate the directives and Wh-words in the directives from elementary 2nd, 4th and 6th grade Korean textbooks. After entering all directives into Microsoft Office Excel, directives with Wh-words were separated. The analysis program, Natmal, was used for the analysis of the directives and Wh-words. The criteria from previous studies were also applied for this analysis process. As a result of the study, there are a lot of nouns and verbs in directives. They were consisted of sentences with an average of 6.9 Eojeol. There were a total of 11 types of Wh-words and 'Mueot(what), Eotteon(which), eotteohge(how)' appeared most frequently in all grades. For question types, both grades had more inferential questions than literal information questions. This results were expected to be used as basic data for language interventions with school aged children who have language disorders.