• KIISE Transactions on Computing Practices
• /
• v.23 no.4
• /
• pp.268-273
• /
• 2017

Recently, deep learning encoder-based approach has been actively applied in the field of sentiment classification. However, Long Short-Term Memory network deep learning encoder, the commonly used architecture, lacks the quality of vector representation when the length of the documents is prolonged. In this study, for effective classification of the sentiment documents, we suggest the use of attention method-based deep learning encoder that generates document vector representation by weighted sum of the outputs of Long Short-Term Memory network based on importance. In addition, we propose methods to modify the attention method-based deep learning encoder to suit the sentiment classification field, which consist of a part that is to applied to window attention method and an attention weight adjustment part. In the window attention method part, the weights are obtained in the window units to effectively recognize feeling features that consist of more than one word. In the attention weight adjustment part, the learned weights are smoothened. Experimental results revealed that the performance of the proposed method outperformed Long Short-Term Memory network encoder, showing 89.67% in accuracy criteria.

• Journal of Intelligence and Information Systems
• /
• v.21 no.1
• /
• pp.1-13
• /
• 2015

As opinion mining in big data applications has been highlighted, a lot of research on unstructured data has made. Lots of social media on the Internet generate unstructured or semi-structured data every second and they are often made by natural or human languages we use in daily life. Many words in human languages have multiple meanings or senses. In this result, it is very difficult for computers to extract useful information from these datasets. Traditional web search engines are usually based on keyword search, resulting in incorrect search results which are far from users' intentions. Even though a lot of progress in enhancing the performance of search engines has made over the last years in order to provide users with appropriate results, there is still so much to improve it. Word sense disambiguation can play a very important role in dealing with natural language processing and is considered as one of the most difficult problems in this area. Major approaches to word sense disambiguation can be classified as knowledge-base, supervised corpus-based, and unsupervised corpus-based approaches. This paper presents a method which automatically generates a corpus for word sense disambiguation by taking advantage of examples in existing dictionaries and avoids expensive sense tagging processes. It experiments the effectiveness of the method based on Naïve Bayes Model, which is one of supervised learning algorithms, by using Korean standard unabridged dictionary and Sejong Corpus. Korean standard unabridged dictionary has approximately 57,000 sentences. Sejong Corpus has about 790,000 sentences tagged with part-of-speech and senses all together. For the experiment of this study, Korean standard unabridged dictionary and Sejong Corpus were experimented as a combination and separate entities using cross validation. Only nouns, target subjects in word sense disambiguation, were selected. 93,522 word senses among 265,655 nouns and 56,914 sentences from related proverbs and examples were additionally combined in the corpus. Sejong Corpus was easily merged with Korean standard unabridged dictionary because Sejong Corpus was tagged based on sense indices defined by Korean standard unabridged dictionary. Sense vectors were formed after the merged corpus was created. Terms used in creating sense vectors were added in the named entity dictionary of Korean morphological analyzer. By using the extended named entity dictionary, term vectors were extracted from the input sentences and then term vectors for the sentences were created. Given the extracted term vector and the sense vector model made during the pre-processing stage, the sense-tagged terms were determined by the vector space model based word sense disambiguation. In addition, this study shows the effectiveness of merged corpus from examples in Korean standard unabridged dictionary and Sejong Corpus. The experiment shows the better results in precision and recall are found with the merged corpus. This study suggests it can practically enhance the performance of internet search engines and help us to understand more accurate meaning of a sentence in natural language processing pertinent to search engines, opinion mining, and text mining. Naïve Bayes classifier used in this study represents a supervised learning algorithm and uses Bayes theorem. Naïve Bayes classifier has an assumption that all senses are independent. Even though the assumption of Naïve Bayes classifier is not realistic and ignores the correlation between attributes, Naïve Bayes classifier is widely used because of its simplicity and in practice it is known to be very effective in many applications such as text classification and medical diagnosis. However, further research need to be carried out to consider all possible combinations and/or partial combinations of all senses in a sentence. Also, the effectiveness of word sense disambiguation may be improved if rhetorical structures or morphological dependencies between words are analyzed through syntactic analysis.

• Journal of Intelligence and Information Systems
• /
• v.27 no.2
• /
• pp.17-32
• /
• 2021

Speaker recognition is generally divided into speaker identification and speaker verification. Speaker recognition plays an important function in the automatic voice system, and the importance of speaker recognition technology is becoming more prominent as the recent development of portable devices, voice technology, and audio content fields continue to expand. Previous speaker recognition studies have been conducted with the goal of automatically determining who the speaker is based on voice files and improving accuracy. Speech is an important sociolinguistic subject, and it contains very useful information that reveals the speaker's attitude, conversation intention, and personality, and this can be an important clue to speaker recognition. The final ending used in the speaker's speech determines the type of sentence or has functions and information such as the speaker's intention, psychological attitude, or relationship to the listener. The use of the terminating ending has various probabilities depending on the characteristics of the speaker, so the type and distribution of the terminating ending of a specific unidentified speaker will be helpful in recognizing the speaker. However, there have been few studies that considered speech in the existing text-based speaker recognition, and if speech information is added to the speech signal-based speaker recognition technique, the accuracy of speaker recognition can be further improved. Hence, the purpose of this paper is to propose a novel method using speech style expressed as a sentence-final ending to improve the accuracy of Korean speaker recognition. To this end, a method called sentence sequencing that generates vector values by using the type and frequency of the sentence-final ending appearing in the utterance of a specific person is proposed. To evaluate the performance of the proposed method, learning and performance evaluation were conducted with a actual drama script. The method proposed in this study can be used as a means to improve the performance of Korean speech recognition service.

• Journal of KIISE
• /
• v.42 no.3
• /
• pp.367-378
• /
• 2015

This study was directed at the design of a hybrid algorithm for competition relation extraction. Previous works on relation extraction have relied on various lexical and deep parsing indicators and mostly utilize only the machine learning method. We present a new algorithm integrating machine learning with various filtering methods. Some simple but useful features for competition relation extraction are also introduced, and an optimum feature set is proposed. The goal of this paper was to increase the precision of competition relation extraction by combining supervised learning with various filtering methods. Filtering methods were employed for classifying compete relation occurrence, using distance restriction for the filtering of feature pairs, and classifying whether or not the candidate entity pair is spam. For evaluation, a test set consisting of 2,565 sentences was examined. The proposed method was compared with the rule-based method and general relation extraction method. As a result, the rule-based method achieved positive precision of 0.812 and accuracy of 0.568, while the general relation extraction method achieved 0.612 and 0.563, respectively. The proposed system obtained positive precision of 0.922 and accuracy of 0.713. These results demonstrate that the developed method is effective for competition relation extraction.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.115-118
• /
• 2002

Parameters used in a speaker recognition system are desirable expressing speaker's characteristics filly and have in a speech. That is to say, if inter-speaker than intra-speaker variance a big characteristic, it is useful to distinguish between speakers. Also, to make minimum error between speakers, it is required the improved recognition technology as well as the distinguishing characteristics. When we see the result of recent simulation performance, we obtain more exact performance by using dynamic characteristics and constant characteristics by a speaking habit. Therefore we suggest it to solve this problem as followings. The prosodic information is used by a characteristic vector of speech. Characteristics vector generally using in speaker recognition system is a modeling spectrum information and is working for a high performance in non-noise circumstance. However, it is found a problem that characteristic vector is distorted in noise circumstance and it makes a reduction of recognition rate. In this paper, we change pitch line divided by segment which can estimate a dynamic characteristic and it is used as a recognition characteristic. we confirmed that the dynamic characteristic is very robust in noise circumstance with a simulation. We make a decision of acceptance or rejection by comparing test pattern and recognition rate using the proposed algorithm has more improvement than using spectrum and prosodic information. Especially stational recognition rate can be obtained in noise circumstance through the simulation.

• 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.

• Journal of Platform Technology
• /
• v.9 no.3
• /
• pp.52-62
• /
• 2021

Quality control is critical at manufacturing sites and is key to predicting the risk of quality defect before manufacturing. However, the reliability of manual quality control methods is affected by human and physical limitations because manufacturing processes vary across industries. These limitations become particularly obvious in domain areas with numerous manufacturing processes, such as the manufacture of major nuclear equipment. This study proposed a novel method for predicting the risk of quality defects by using natural language processing and machine learning. In this study, production data collected over 6 years at a factory that manufactures main equipment that is installed in nuclear power plants were used. In the preprocessing stage of text data, a mapping method was applied to the word dictionary so that domain knowledge could be appropriately reflected, and a hybrid algorithm, which combined n-gram, Term Frequency-Inverse Document Frequency, and Singular Value Decomposition, was constructed for sentence vectorization. Next, in the experiment to classify the risky processes resulting in poor quality, k-fold cross-validation was applied to categorize cases from Unigram to cumulative Trigram. Furthermore, for achieving objective experimental results, Naive Bayes and Support Vector Machine were used as classification algorithms and the maximum accuracy and F1-score of 0.7685 and 0.8641, respectively, were achieved. Thus, the proposed method is effective. The performance of the proposed method were compared and with votes of field engineers, and the results revealed that the proposed method outperformed field engineers. Thus, the method can be implemented for quality control at manufacturing sites.

• Journal of KIISE
• /
• v.43 no.11
• /
• pp.1210-1215
• /
• 2016

Readership of online newspapers has grown with the proliferation of smart devices. However, fierce competition between Internet newspaper companies has resulted in a large increase in the number of hoax articles. Hoax articles are those where the title does not convey the content of the main story, and this gives readers the wrong information about the contents. We note that the hoax articles have certain characteristics, such as unnecessary celebrity quotations, mismatch in the title and content, or incomplete sentences. Based on these, we extract and validate features to identify hoax articles. We build a large-scale training dataset by analyzing text keywords in replies to articles and thus extracted five effective features. We evaluate the performance of the support vector machine classifier on the extracted features, and a 92% accuracy is observed in our validation set. In addition, we also present a selective bigram model to measure the consistency between the title and content, which can be effectively used to analyze short texts in general.

• 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 Korean Institute of Intelligent Systems
• /
• v.14 no.7
• /
• pp.895-901
• /
• 2004

This paper describes a two-phase intelligent method for filtering spam mail based on textual information and hyperlinks. Scince the body of spam mail has little text information, it provides insufficient hints to distinguish spam mails from legitimate mails. To resolve this problem, we follows hyperlinks contained in the email body, fetches contents of a remote webpage, and extracts hints (i.e., features) from original email body and fetched webpages. We divided hints into two kinds of information: definite information (sender`s information and definite spam keyword lists) and less definite textual information (words or phrases, and particular features of email). In filtering spam mails, definite information is used first, and then less definite textual information is applied. In our experiment, the method of fetching web pages achieved an improvement of F-measure by 9.4% over the method of using on original email header and body only.