• Phonetics and Speech Sciences
• /
• v.6 no.3
• /
• pp.155-164
• /
• 2014

This paper proposes a new method to determine the recognition units for large vocabulary continuous speech recognition (LVCSR) in Korean by applying unsupervised segmentation and merging. In the proposed method, a text sentence is segmented into morphemes and position information is added to morphemes. Then submorpheme units are obtained by splitting the morpheme units through the maximization of posterior probability terms. The posterior probability terms are computed from the morpheme frequency distribution, the morpheme length distribution, and the morpheme frequency-of-frequency distribution. Finally, the recognition units are obtained by sequentially merging the submorpheme pair with the highest frequency. Computer experiments are conducted using a Korean LVCSR with a 100k word vocabulary and a trigram language model obtained by a 300 million eojeol (word phrase) corpus. The proposed method is shown to reduce the out-of-vocabulary rate to 1.8% and reduce the syllable error rate relatively by 14.0%.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.4
• /
• pp.407-414
• /
• 2002

In this paper, we present a method that automatically generates concatenate morpheme based language models to improve the performance of Korean large vocabulary continuous speech recognition. The focus was brought into improvement against recognition errors of monosyllable morphemes that occupy 54% of the training text corpus and more frequently mis-recognized. Knowledge-based method using POS patterns has disadvantages such as the difficulty in making rules and producing many low frequency concatenate morphemes. Proposed method automatically selects morpheme-pairs from training text data based on measures such as frequency, mutual information, and unigram log likelihood. Experiment was performed using 7M-morpheme text corpus and 20K-morpheme lexicon. The frequency measure with constraint on the number of morphemes used for concatenation produces the best result of reducing monosyllables from 54% to 30%, bigram perplexity from 117.9 to 97.3. and MER from 21.3% to 17.6%.

• Journal of KIISE:Software and Applications
• /
• v.33 no.11
• /
• pp.965-978
• /
• 2006

This paper implements an automatic Korean word-spacing system based on word-recognition using morpheme unigrams and the pattern that the categories of those morpheme unigrams share within a candidate word. Although previous work on Korean word-spacing models has produced the advantages of easy construction and time efficiency, there still remain problems, such as data sparseness and critical memory size, which arise from the morpho-typological characteristics of Korean. In order to cope with both problems, our implementation uses the stochastic information of morpheme unigrams, and their category patterns, instead of word unigrams. A word's probability in a sentence is obtained based on morpheme probability and the weight for the morpheme's category within the category pattern of the candidate word. The category weights are trained so as to minimize the error means between the observed probabilities of words and those estimated by words' individual-morphemes' probabilities weighted according to their categories' powers in a given word's category pattern.

• Language and Information
• /
• v.18 no.1
• /
• pp.1-32
• /
• 2014

This paper provides a corpus study on subjunctives in Korean in a way of comparative semantics. The whole arguments of this paper are bolstered by distributional evidence taken from naturally occurring bitexts (i.e. a bilingual corpus), in which one sentence in a language is aligned with one translation in the other language. Since previous studies regard past tense morphology as the main component to express irrealis and uncertainty, this paper accordingly checks out whether the past tense morpheme (e/a)ss in Korean is also responsible for conveying the meaning of subjunctives. My finding is that the past tense morpheme (e/a)ss is a sufficient condition for forming subjunctives in Korean. The current corpus study verifies that the past tense morpheme is not obligatorily used in present conditional counterfactuals in Korean, unlike English. Yet, if (e/a)ss is used and the antecedent denotes a present situation, the conditional sentence can only be interpreted as conveying counterfactuality. On the other hand, wish constructions in Korean, irrespective of the semantic tense, often contain the past tense morpheme. Hence, this work substantiates Iatridou (2000)'s theory of 'fake past tense' is applicable to Korean subjunctives. The present corpus study, additionally, reveals that a conditional marker telamyen is a component of expressing past counterfactuals in Korean.

• Journal of Information Technology Services
• /
• v.17 no.1
• /
• pp.79-89
• /
• 2018

Sentiment analysis is a technique of text mining that extracts feelings of the person who wrote the sentence like movie review. The preliminary researches of sentiment analysis identify sentiments by using the dictionary which contains negative and positive words collected in advance. As researches on deep learning are actively carried out, sentiment analysis using deep learning model with morpheme or word unit has been done. However, this model has disadvantages in that the word dictionary varies according to the domain and the number of morphemes or words gets relatively larger than that of phonemes. Therefore, the size of the dictionary becomes large and the complexity of the model increases accordingly. We construct a sentiment analysis model using recurrent neural network by dividing input data into phoneme-level which is smaller than morpheme-level. To verify the performance, we use 30,000 movie reviews from the Korean biggest portal, Naver. Morpheme-level sentiment analysis model is also implemented and compared. As a result, the phoneme-level sentiment analysis model is superior to that of the morpheme-level, and in particular, the phoneme-level model using LSTM performs better than that of using GRU model. It is expected that Korean text processing based on a phoneme-level model can be applied to various text mining and language models.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.4
• /
• pp.17-22
• /
• 1999

It is difficult to integrate the speech processing systems and machine translation system in the spoken language translation system by reason that each system uses its own data and basic processing unit. So, we need a common I/O unit which is used in the whole system. In this paper, we propose a Pscudo-Morpheme as the interface between speech processing systems and language translation system. We implement a morphological analysis system for Pseudo-morpheme. The speech processing system using this pseudo-morpheme can get better result than other systems using the phrase or the general morpheme. So, the quality of the whole spoken language translation system can be improved. The analysis-ratio of our implemented system is 98.9%. This is similar to the common morphological analysis systems.

• The Journal of the Acoustical Society of Korea
• /
• v.15 no.6
• /
• pp.90-98
• /
• 1996

In this paper we have proposed the sign language transformation system for deaf based on a morpheme analysis. The proposed system extracts phoneme components and connection informations of the input character sequence by using a morpheme analysis. And then the sign image obtained by component analysis is correctly and automatically generated through the sign image database. For the effective sign language transformation, the language description dictionary which consists of a morpheme analysis part for analysis of input character sequence and sign language description part for reference of sign language pattern is costructed. To avoid the duplicating sign language pattern, the pattern is classified a basic, a compound and a similar sign word. The computer simulation shows the usefulness of the proposed system.

• Speech Sciences
• /
• v.9 no.4
• /
• pp.77-91
• /
• 2002

This study investigated the use of grammatical morphemes (substantive morphemes and connective endings) of Korean speaking children with and without language impairment (LI). Participants were two children (ages 5;11 and 6;2) with SLI (specific language impairment), two LD (language delay) children (ages 6;3 and 6;5) with 70-84 range on a performance-IQ test, and two children (ages 5;7 and 6;1) with ND (normal development). Spontaneous language samples were elicited by play activities and story generation. A total of 8,059 (M=I,343, ranged 966-1,659) intelligible and nonimitative utterances were analyzed by the KCLA 2.0 (Korean Computerized Language Analysis 2.0) program for substantive morphemes and connective endings. The findings of this study were as follows; (1) The Korean speaking children with LI including SLI demonstrated less uses of grammatical morphemes than ND children. (2) Few differences were found between LI and ND children in the use of the grammatical morpheme types. (3) LI children produced significantly higher percentage of grammatical morpheme errors in spontaneous speech than ND children. (4) Few differences were found between SLI and LD children in degrees of the use and the error of grammatical morphemes.

• The KIPS Transactions:PartB
• /
• v.19B no.3
• /
• pp.195-200
• /
• 2012

In Korean, spelling change in various forms must be recovered into base forms in morphological analysis as well as part-of-speech (POS) tagging is difficult without morphological analysis because Korean is agglutinative. This is one of notorious problems in Korean morphological analysis and has been solved by morpheme recovery rules, which generate morphological ambiguity resolved by POS tagging. In this paper, we propose a morpheme recovery scheme based on machine learning methods like Na$\ddot{i}$ve Bayes models. Input features of the models are the surrounding context of the syllable which the spelling change is occurred and categories of the models are the recovered syllables. The POS tagging system with the proposed model has demonstrated the $F_1$-score of 97.5% for the ETRI tree-tagged corpus. Thus it can be decided that the proposed model is very useful to handle morpheme recovery in Korean.

• The Journal of the Korea Contents Association
• /
• v.6 no.10
• /
• pp.9-16
• /
• 2006

In this paper, the acoustic model output of CSR(Continuous Speech Recognition) was used to analyze news data News database used in this experiment was consisted of 2,093 articles. Due to the low efficiency of language model, conventional Korean CSR is not appropriate to the analysis of news data. This problem could be handled successfully by introducing post-processing work of recognition result of acoustic model. The acoustic model more robust than language model in Korean environment. The result of post-processing work was made into KIF(Keyword information file). When threshold of acoustic model's output level was 100, 86.9% of whole target morpheme was included in post-processing result. At the same condition, applying length information based normalization, 81.25% of whole target morpheme was recognized. The purpose of normalization was to compensate long-length morpheme. According to experiment result, 75.13% of whole target morpheme was recognized KIF(314MB) had been produced from original news data(5,040MB). The decrease rate of absolute information met was approximately 93.8%.