• 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 Institute of Intelligent Transport Systems
• /
• v.16 no.6
• /
• pp.141-155
• /
• 2017

As the use of SNS becomes more active, many people are posting their thoughts about specific events in their SNS in the form of text. As a result, SNS is used in various fields such as finance and distribution to conduct service satisfaction surveys and consumer monitoring. However, in the transportation area, there are not enough cases to utilize unstructured data analysis such as emotional analysis. In this study, we developed an emotional analysis methodology that can be used in transportation by using highway VOC data, which is atypical data collected by Korea Expressway Corporation. The developed methodology consists of morpheme analysis, emotional dictionary construction, and emotional discrimination of the collected unstructured data. The developed methodology was verified using highway related tweet data. As a result of the analysis, it can be guessed that many information and information about the construction and the accident were related to the highway during the analysis period. Also, it seems that users complain about the delay caused by construction and accident.

• Journal of Korean Society of Archives and Records Management
• /
• v.17 no.2
• /
• pp.165-188
• /
• 2017

As the recognition of the importance of user-oriented services is increasing, there has been a heightened attention for finding aids that could improve the effectiveness of searching. This study tried to draw various facet elements that can be applied to the presidential records retrieval system using presidential records as cases in analyzing various resources, considering the importance of facets in finding aids for the improvement of effectiveness in searching in the future and the importance of presidential records in Korea. In drawing facet elements based on the characteristics of presidential records, the websites of the National Archives (NARA) and Presidential (Prime Ministers') Archives as well as their search options were examined as cases. In addition, the morpheme of each title of presidential records were analyzed, as well as the terms entered by the users of the Presidential Archives Portal of Korea, the terms used in the request for information disclosure toward the Presidential Archives in Korea, the search options of the Presidential Archives Portal, and the elements of the description and metadata standards. The significance of this study lies on suggesting the methodology of developing various facets as main elements in finding aids using the presidential records as cases.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.11
• /
• pp.555-562
• /
• 2016

Semantic Role Labeling (SRL) is to determine the semantic relation of a predicate and its argu-ments in a sentence. But Korean semantic role labeling has faced on difficulty due to its different language structure compared to English, which makes it very hard to use appropriate approaches developed so far. That means that methods proposed so far could not show a satisfied perfor-mance, compared to English and Chinese. To complement these problems, we focus on suffix information analysis, such as josa (case suffix) and eomi (verbal ending) analysis. Korean lan-guage is one of the agglutinative languages, such as Japanese, which have well defined suffix structure in their words. The agglutinative languages could have free word order due to its de-veloped suffix structure. Also arguments with a single morpheme are then labeled with statistics. In addition, machine learning algorithms such as Support Vector Machine (SVM) and Condi-tional Random Fields (CRF) are used to model SRL problem on arguments that are not labeled at the suffix analysis phase. The proposed method is intended to reduce the range of argument instances to which machine learning approaches should be applied, resulting in uncertain and inaccurate role labeling. In experiments, we use 15,224 arguments and we are able to obtain approximately 83.24% f1-score, increased about 4.85% points compared to the state-of-the-art Korean SRL research.

• Smart Media Journal
• /
• v.12 no.11
• /
• pp.185-197
• /
• 2023

This study was conducted using text mining and network theory to extract useful information for application for occupancy and performance of permit tasks contained in the permit contents from the permit register, which is used only for the simple purpose of recording occupancy permit information. Based on text mining, we analyzed and compared the frequency of vocabulary occurrence and topic modeling in five regions, including Seoul, Gyeonggi, Gyeongsang, Jeolla, Chungcheong, and Gangwon, as well as normalization processes such as stopword removal and morpheme analysis. By applying four types of centrality algorithms, including stage, proximity, mediation, and eigenvector, which are widely used in network theory, we looked at keywords that are in a central position or act as an intermediary in the network. Through a comprehensive analysis of vocabulary appearance frequency, topic modeling, and network centrality, it was found that the 'installation' keyword was the most influential in all regions. This is believed to be the result of the Ministry of Environment's permit management office issuing many permits for constructing facilities or installing structures. In addition, it was found that keywords related to road facilities, flood control facilities, underground facilities, power/communication facilities, sports/park facilities, etc. were at a central position or played a role as an intermediary in topic modeling and networks. Most of the keywords appeared to have a Zipf's law statistical distribution with low frequency of occurrence and low distribution ratio.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.71-88
• /
• 2017

Language models were originally developed for speech recognition and language processing. Using a set of example sentences, a language model predicts the next word or character based on sequential input data. N-gram models have been widely used but this model cannot model the correlation between the input units efficiently since it is a probabilistic model which are based on the frequency of each unit in the training set. Recently, as the deep learning algorithm has been developed, a recurrent neural network (RNN) model and a long short-term memory (LSTM) model have been widely used for the neural language model (Ahn, 2016; Kim et al., 2016; Lee et al., 2016). These models can reflect dependency between the objects that are entered sequentially into the model (Gers and Schmidhuber, 2001; Mikolov et al., 2010; Sundermeyer et al., 2012). In order to learning the neural language model, texts need to be decomposed into words or morphemes. Since, however, a training set of sentences includes a huge number of words or morphemes in general, the size of dictionary is very large and so it increases model complexity. In addition, word-level or morpheme-level models are able to generate vocabularies only which are contained in the training set. Furthermore, with highly morphological languages such as Turkish, Hungarian, Russian, Finnish or Korean, morpheme analyzers have more chance to cause errors in decomposition process (Lankinen et al., 2016). Therefore, this paper proposes a phoneme-level language model for Korean language based on LSTM models. A phoneme such as a vowel or a consonant is the smallest unit that comprises Korean texts. We construct the language model using three or four LSTM layers. Each model was trained using Stochastic Gradient Algorithm and more advanced optimization algorithms such as Adagrad, RMSprop, Adadelta, Adam, Adamax, and Nadam. Simulation study was done with Old Testament texts using a deep learning package Keras based the Theano. After pre-processing the texts, the dataset included 74 of unique characters including vowels, consonants, and punctuation marks. Then we constructed an input vector with 20 consecutive characters and an output with a following 21st character. Finally, total 1,023,411 sets of input-output vectors were included in the dataset and we divided them into training, validation, testsets with proportion 70:15:15. All the simulation were conducted on a system equipped with an Intel Xeon CPU (16 cores) and a NVIDIA GeForce GTX 1080 GPU. We compared the loss function evaluated for the validation set, the perplexity evaluated for the test set, and the time to be taken for training each model. As a result, all the optimization algorithms but the stochastic gradient algorithm showed similar validation loss and perplexity, which are clearly superior to those of the stochastic gradient algorithm. The stochastic gradient algorithm took the longest time to be trained for both 3- and 4-LSTM models. On average, the 4-LSTM layer model took 69% longer training time than the 3-LSTM layer model. However, the validation loss and perplexity were not improved significantly or became even worse for specific conditions. On the other hand, when comparing the automatically generated sentences, the 4-LSTM layer model tended to generate the sentences which are closer to the natural language than the 3-LSTM model. Although there were slight differences in the completeness of the generated sentences between the models, the sentence generation performance was quite satisfactory in any simulation conditions: they generated only legitimate Korean letters and the use of postposition and the conjugation of verbs were almost perfect in the sense of grammar. The results of this study are expected to be widely used for the processing of Korean language in the field of language processing and speech recognition, which are the basis of artificial intelligence systems.

• Phonetics and Speech Sciences
• /
• v.12 no.2
• /
• pp.63-72
• /
• 2020

This study aims to develop a video-based smart utterance deep analyser (SUDA) application that analyzes semiautomatically the utterances that child and mother produce during interactions over time. SUDA runs on the platform of Android, iPhones, and tablet PCs, and allows video recording and uploading to server. In this device, user modes are divided into three modes: expert mode, general mode and manager mode. In the expert mode which is useful for speech and language evaluation, the subject's utterances are analyzed semi-automatically by measuring speech and language factors such as disfluency, morpheme, syllable, word, articulation rate and response time, etc. In the general mode, the outcome of utterance analysis is provided in a graph form, and the manger mode is accessed only to the administrator controlling the entire system, such as utterance analysis and video deletion. SUDA helps to reduce clinicians' and researchers' work burden by saving time for utterance analysis. It also helps parents to receive detailed information about speech and language development of their child easily. Further, this device will contribute to building a big longitudinal data enough to explore predictors of stuttering recovery and persistence.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.11
• /
• pp.441-448
• /
• 2019

Previous researches on automatic spacing of Korean sentences has been researched to correct spacing errors by using n-gram based statistical techniques or morpheme analyzer to insert blanks in the word boundary. In this paper, we propose an end-to-end automatic word spacing by using deep neural network. Automatic word spacing problem could be defined as a tag classification problem in unit of syllable other than word. For contextual representation between syllables, Bi-LSTM encodes the dependency relationship between syllables into a fixed-length vector of continuous vector space using forward and backward LSTM cell. In order to conduct automatic word spacing of Korean sentences, after a fixed-length contextual vector by Bi-LSTM is classified into auto-spacing tag(B or I), the blank is inserted in the front of B tag. For tag classification method, we compose three types of classification neural networks. One is feedforward neural network, another is neural network language model and the other is linear-chain CRF. To compare our models, we measure the performance of automatic word spacing depending on the three of classification networks. linear-chain CRF of them used as classification neural network shows better performance than other models. We used KCC150 corpus as a training and testing data.

• Journal of Intelligence and Information Systems
• /
• v.24 no.4
• /
• pp.111-136
• /
• 2018

In this paper, we propose a methodology to extract answer information about queries from various types of unstructured documents collected from multi-sources existing on web in order to expand knowledge base. The proposed methodology is divided into the following steps. 1) Collect relevant documents from Wikipedia, Naver encyclopedia, and Naver news sources for "subject-predicate" separated queries and classify the proper documents. 2) Determine whether the sentence is suitable for extracting information and derive the confidence. 3) Based on the predicate feature, extract the information in the proper sentence and derive the overall confidence of the information extraction result. In order to evaluate the performance of the information extraction system, we selected 400 queries from the artificial intelligence speaker of SK-Telecom. Compared with the baseline model, it is confirmed that it shows higher performance index than the existing model. The contribution of this study is that we develop a sequence tagging model based on bi-directional LSTM-CRF using the predicate feature of the query, with this we developed a robust model that can maintain high recall performance even in various types of unstructured documents collected from multiple sources. The problem of information extraction for knowledge base extension should take into account heterogeneous characteristics of source-specific document types. The proposed methodology proved to extract information effectively from various types of unstructured documents compared to the baseline model. There is a limitation in previous research that the performance is poor when extracting information about the document type that is different from the training data. In addition, this study can prevent unnecessary information extraction attempts from the documents that do not include the answer information through the process for predicting the suitability of information extraction of documents and sentences before the information extraction step. It is meaningful that we provided a method that precision performance can be maintained even in actual web environment. The information extraction problem for the knowledge base expansion has the characteristic that it can not guarantee whether the document includes the correct answer because it is aimed at the unstructured document existing in the real web. When the question answering is performed on a real web, previous machine reading comprehension studies has a limitation that it shows a low level of precision because it frequently attempts to extract an answer even in a document in which there is no correct answer. The policy that predicts the suitability of document and sentence information extraction is meaningful in that it contributes to maintaining the performance of information extraction even in real web environment. The limitations of this study and future research directions are as follows. First, it is a problem related to data preprocessing. In this study, the unit of knowledge extraction is classified through the morphological analysis based on the open source Konlpy python package, and the information extraction result can be improperly performed because morphological analysis is not performed properly. To enhance the performance of information extraction results, it is necessary to develop an advanced morpheme analyzer. Second, it is a problem of entity ambiguity. The information extraction system of this study can not distinguish the same name that has different intention. If several people with the same name appear in the news, the system may not extract information about the intended query. In future research, it is necessary to take measures to identify the person with the same name. Third, it is a problem of evaluation query data. In this study, we selected 400 of user queries collected from SK Telecom 's interactive artificial intelligent speaker to evaluate the performance of the information extraction system. n this study, we developed evaluation data set using 800 documents (400 questions * 7 articles per question (1 Wikipedia, 3 Naver encyclopedia, 3 Naver news) by judging whether a correct answer is included or not. To ensure the external validity of the study, it is desirable to use more queries to determine the performance of the system. This is a costly activity that must be done manually. Future research needs to evaluate the system for more queries. It is also necessary to develop a Korean benchmark data set of information extraction system for queries from multi-source web documents to build an environment that can evaluate the results more objectively.

• Journal of Intelligence and Information Systems
• /
• v.18 no.2
• /
• pp.143-156
• /
• 2012

People easily believe that news and stock index are closely related. They think that securing news before anyone else can help them forecast the stock prices and enjoy great profit, or perhaps capture the investment opportunity. However, it is no easy feat to determine to what extent the two are related, come up with the investment decision based on news, or find out such investment information is valid. If the significance of news and its impact on the stock market are analyzed, it will be possible to extract the information that can assist the investment decisions. The reality however is that the world is inundated with a massive wave of news in real time. And news is not patterned text. This study suggests the stock-index invest model based on "News Big Data" opinion mining that systematically collects, categorizes and analyzes the news and creates investment information. To verify the validity of the model, the relationship between the result of news opinion mining and stock-index was empirically analyzed by using statistics. Steps in the mining that converts news into information for investment decision making, are as follows. First, it is indexing information of news after getting a supply of news from news provider that collects news on real-time basis. Not only contents of news but also various information such as media, time, and news type and so on are collected and classified, and then are reworked as variable from which investment decision making can be inferred. Next step is to derive word that can judge polarity by separating text of news contents into morpheme, and to tag positive/negative polarity of each word by comparing this with sentimental dictionary. Third, positive/negative polarity of news is judged by using indexed classification information and scoring rule, and then final investment decision making information is derived according to daily scoring criteria. For this study, KOSPI index and its fluctuation range has been collected for 63 days that stock market was open during 3 months from July 2011 to September in Korea Exchange, and news data was collected by parsing 766 articles of economic news media M company on web page among article carried on stock information>news>main news of portal site Naver.com. In change of the price index of stocks during 3 months, it rose on 33 days and fell on 30 days, and news contents included 197 news articles before opening of stock market, 385 news articles during the session, 184 news articles after closing of market. Results of mining of collected news contents and of comparison with stock price showed that positive/negative opinion of news contents had significant relation with stock price, and change of the price index of stocks could be better explained in case of applying news opinion by deriving in positive/negative ratio instead of judging between simplified positive and negative opinion. And in order to check whether news had an effect on fluctuation of stock price, or at least went ahead of fluctuation of stock price, in the results that change of stock price was compared only with news happening before opening of stock market, it was verified to be statistically significant as well. In addition, because news contained various type and information such as social, economic, and overseas news, and corporate earnings, the present condition of type of industry, market outlook, the present condition of market and so on, it was expected that influence on stock market or significance of the relation would be different according to the type of news, and therefore each type of news was compared with fluctuation of stock price, and the results showed that market condition, outlook, and overseas news was the most useful to explain fluctuation of news. On the contrary, news about individual company was not statistically significant, but opinion mining value showed tendency opposite to stock price, and the reason can be thought to be the appearance of promotional and planned news for preventing stock price from falling. Finally, multiple regression analysis and logistic regression analysis was carried out in order to derive function of investment decision making on the basis of relation between positive/negative opinion of news and stock price, and the results showed that regression equation using variable of market conditions, outlook, and overseas news before opening of stock market was statistically significant, and classification accuracy of logistic regression accuracy results was shown to be 70.0% in rise of stock price, 78.8% in fall of stock price, and 74.6% on average. This study first analyzed relation between news and stock price through analyzing and quantifying sensitivity of atypical news contents by using opinion mining among big data analysis techniques, and furthermore, proposed and verified smart investment decision making model that could systematically carry out opinion mining and derive and support investment information. This shows that news can be used as variable to predict the price index of stocks for investment, and it is expected the model can be used as real investment support system if it is implemented as system and verified in the future.