• Journal of the Korea Society of Computer and Information
• /
• v.28 no.9
• /
• pp.35-46
• /
• 2023

In this study, we propose a Dual Approach methodology to enhance the accuracy of document classifiers by utilizing both contextual and keyword information. Firstly, contextual information is extracted using Google's BERT, a pre-trained language model known for its outstanding performance in various natural language understanding tasks. Specifically, we employ KoBERT, a pre-trained model on the Korean corpus, to extract contextual information in the form of the CLS token. Secondly, keyword information is generated for each document by encoding the set of keywords into a single vector using an Autoencoder. We applied the proposed approach to 40,130 documents related to healthcare and medicine from the National R&D Projects database of the National Science and Technology Information Service (NTIS). The experimental results demonstrate that the proposed methodology outperforms existing methods that rely solely on document or word information in terms of accuracy for document classification.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.3
• /
• pp.330-338
• /
• 2002

Spontaneous speech is ungrammatical as well as serious phonological variations, which make recognition extremely difficult, compared with read speech. In this paper, for conversational speech recognition, we analyze the transcriptions of the real conversational speech, and then classify the characteristics of conversational speech in the speech recognition aspect. Reflecting these features, we obtain the baseline system for conversational speech recognition. The classification consists of long duration of silence, disfluencies and phonological variations; each of them is classified with similar features. To deal with these characteristics, first, we update silence model and append a filled pause model, a garbage model; second, we append multiple phonetic transcriptions to lexicon for most frequent phonological variations. In our experiments, our baseline morpheme error rate (WER) is 31.65%; we obtain MER reductions such as 2.08% for silence and garbage model, 0.73% for filled pause model, and 0.73% for phonological variations. Finally, we obtain 27.92% MER for conversational speech recognition, which will be used as a baseline for further study.

• Annual Conference on Human and Language Technology
• /
• 2012.10a
• /
• pp.107-112
• /
• 2012

본 논문은 일반 텍스트에 나타나는 경쟁 관계에 있는 고유명사들을 경쟁자로 자동 추출하는 방법에 대한 것으로, 규칙 기반 방법과 기계 학습 기반 방법을 모두 제안하고 비교하였다. 제안한 시스템은 뉴스 기사를 대상으로 하였고, 문장에 경쟁관계를 나타내는 명확한 정보가 있는 경우에만 추출하는 것을 목표로 하였다. 규칙기반 경쟁어 추출 시스템은 2개의 고유명사가 경쟁관계임을 나타내는 단서단어에 기반해서 경쟁어를 추출하는 시스템이며, 경쟁표현 단서단어는 620개가 수집되어 사용됐다. 기계학습 기반 경쟁어 추출시스템은 경쟁어 추출을 경쟁어 후보에 대한 경쟁여부의 바이너리 분류 문제로 접근하였다. 분류 알고리즘은 Support Vector Machines을 사용하였고, 경쟁어 주변 문맥 정보를 대표할 수 있는 언어 독립적 5개 자질에 기반해서 모델을 학습하였다. 성능평가를 위해서 이슈화되고 있는 핫키워드 54개에 대해서 623개의 경쟁어를 뉴스 기사로부터 수집해서 평가셋을 구축하였다. 비교 평가를 위해서 기준시스템으로 연관어에 기반해서 경쟁어를 추출하는 시스템을 구현하였고, Recall/Precision/F1 성능으로 0.119/0.214/0.153을 얻었다. 제안 시스템의 실험 결과로 규칙기반 시스템은 0.793/0.207/0.328 성능을 보였고, 기계 학습기반 시스템은 0.578/0.730/0.645 성능을 보였다. Recall 성능은 규칙기반 시스템이 0.793으로 가장 좋았고, 기준시스템에 비해서 67.4%의 성능 향상이 있었다. Precision과 F1 성능은 기계학습기반 시스템이 0.730과 0.645로 가장 좋았고, 기준시스템에 비해서 각각 61.6%, 49.2%의 성능향상이 있었다. 기준시스템에 비해서 제안한 시스템이 Recall, Precision, F1 성능이 모두 대폭적으로 향상되었으므로 제안한 방법이 효과적임을 알 수 있다.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.11
• /
• pp.501-512
• /
• 2021

Recently, a breakthrough has been made in the NLP area by Transformer techniques based on encoder-decoder. However, this only can be used in mainstream languages where millions of dataset are well-equipped, such as English and Chinese, and there is a limitation that it cannot be used in non-mainstream languages where dataset are not established. In addition, there is a deflection problem that focuses on the beginning of the document in mechanical summarization. Therefore, these methods are not suitable for documents with flows such as fairy tales and novels. In this paper, we propose a hybrid summarization method that does not require a dataset and improves the deflection problem using GAN with two adaptive discriminators. We evaluate our model on the CNN/Daily Mail dataset to verify an objective validity. Also, we proved that the model has valid performance in Korean, one of the non-mainstream languages.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.12
• /
• pp.3052-3062
• /
• 1997

An Information Retrieval (IR) is to retrieve relevant information that satisfies user's information needs. However a major role of IR systems is not just the generation of sets of relevant documents, but to help determine which documents are most likely to be relevant to the given requirements. Various attempts have been made in the recent past to use syntactic analysis methods for the generation of complex construction that are essential for content identification in various automatic text analysis systems. Unfortunately, it is known that methods based on syntactic understanding alone are not sufficiently powerful to Produce complete analyses of arbitrary text samples. In this paper, we present a document ranking method based on two-level ranking. The first level is used to retrieve the documents, and the second level to reorder the retrieved documents. The main keywords used in the first level can be defined as nouns and/or compound nouns that possess good document discrimination powers. The sub-keywords used in the second level can be also defined as adjectives, adverbs, and/or verbs that are not main keywords, and function words. An empirical study was conducted from a Korean encyclopedia with 23,113 entries and 161 Korean natural language queries collected by end users. 850% of the natural language queries contained sub-keywords. The two-level document ranking methods provides significant improvement in retrieval effectiveness over traditional ranking methods.

• Journal of KIISE:Software and Applications
• /
• v.35 no.5
• /
• pp.306-312
• /
• 2008

In a goal-oriented dialogue, speaker's intention can be approximated by a semantic structure that consists of a pair of a speech act and a concept sequence. Therefore, it is very important to correctly identify the semantic structure of an utterance for implementing an intelligent dialogue system. In this paper, we propose a model to efficiently analyze the semantic structures based on an active teaming method. To reduce the burdens of high-level linguistic analysis, the proposed model only uses morphological features and previous semantic structures as input features. To improve the precisions of semantic structure analysis, the proposed model adopts CRFs(Conditional Random Fields), which show high performances in natural language processing, as an underlying statistical model. In the experiments in a schedule arrangement domain, we found that the proposed model shows similar performances(92.4% in speech act analysis and 89.8% in concept sequence analysis) to the previous models although it uses about a third of training data.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.18 no.4
• /
• pp.501-506
• /
• 2008

Statistical POS tagging is prone to error, because of the inherent limitations of statistical data, especially single source of data. Therefore it is widely agreed that the possibility of further enhancement lies in exploiting various knowledge sources. However these data sources are bound to be inconsistent to each other. This paper shows the possibility of using maximum entropy model to Korean language POS tagging. We use as the knowledge sources n-gram data and trigger pair data. We show how perplexity measure varies when two knowledge sources are combined using maximum entropy method. The experiment used a trigram model which produced 94.9% accuracy using Hidden Markov Model, and showed increase to 95.6% when combined with trigger pair data using Maximum Entropy method. This clearly shows possibility of further enhancement when various knowledge sources are developed and combined using ME method.

• Phonetics and Speech Sciences
• /
• v.12 no.4
• /
• pp.63-71
• /
• 2020

Speech recognition is one of the areas actively commercialized using deep learning and machine learning techniques. However, the majority of speech recognition systems on the market are developed on data with limited diversity of speakers and tend to perform well on typical adult speakers only. This is because most of the speech recognition models are generally learned using a speech database obtained from adult males and females. This tends to cause problems in recognizing the speech of the elderly, children and people with dialects well. To solve these problems, it may be necessary to retain big database or to collect a data for applying a speaker adaptation. However, this paper proposes that a new end-to-end speech recognition method consists of an acoustic augmented recurrent encoder and a transformer decoder with linguistic prediction. The proposed method can bring about the reliable performance of acoustic and language models in limited data conditions. The proposed method was evaluated to recognize Korean elderly and children speech with limited amount of training data and showed the better performance compared of a conventional method.

• Smart Media Journal
• /
• v.13 no.6
• /
• pp.44-53
• /
• 2024

Companies in modern society are increasingly recognizing sentiment classification as a crucial task, emphasizing the importance of accurately understanding consumer opinions opinions across various platforms such as social media, product reviews, and customer feedback for competitive success. Extensive research is being conducted on sentiment classification as it helps improve products or services by identifying the diverse opinions and emotions of consumers. In sentiment classification, fine-tuning with large-scale datasets and pre-trained language models is essential for enhancing performance. Recent advancements in artificial intelligence have led to high-performing sentiment classification models, with the ELECTRA model standing out due to its efficient learning methods and minimal computing resource requirements. Therefore, this paper proposes a method to enhance sentiment classification performance through efficient fine-tuning of various datasets using the KoELECTRA model, specifically trained for Korean.

• Journal of The Korean Association For Science Education
• /
• v.36 no.3
• /
• pp.389-397
• /
• 2016

This study aims to test the efficacy of English-based automated computer scoring models and machine translation to score Korean college students' written responses on natural selection concept items. To this end, I collected 128 pre-service biology teachers' written responses on four-item instrument (total 512 written responses). The machine translation software (i.e., Google Translate) translated both original responses and spell-corrected responses. The presence/absence of five scientific ideas and three $na{\ddot{i}}ve$ ideas in both translated responses were judged by the automated computer scoring models (i.e., EvoGrader). The computer-scored results (4096 predictions) were compared with expert-scored results. The results illustrated that no significant differences in both average scores and statistical results using average scores was found between the computer-scored result and experts-scored result. The Pearson correlation coefficients of composite scores for each student between computer scoring and experts scoring were 0.848 for scientific ideas and 0.776 for $na{\ddot{i}}ve$ ideas. The inter-rater reliability indices (Cohen kappa) between computer scoring and experts scoring for linguistically simple concepts (e.g., variation, competition, and limited resources) were over 0.8. These findings reveal that the English-based automated computer scoring models and machine translation can be a promising method in scoring Korean college students' written responses on natural selection concept items.