• Annual Conference on Human and Language Technology
• /
• 2014.10a
• /
• pp.95-99
• /
• 2014

본 논문에서는 최근 급속히 증가하여 사회적 이슈가 되고 있는 SMS 스팸 필터링을 위한 듀얼 SMS 스팸필터링 기법을 제안한다. 지속적으로 증가하고 새롭게 변형되는 SMS 문자 필터링을 위해서는 패턴 및 스팸 단어 사전을 통한 필터링은 많은 수작업을 요구하여 부적합하다. 그리하여 기계 학습을 이용한 자동화 시스템 구축이 요구되고 있으며, 효과적인 기계 학습을 위해서는 자질 선택과 자질의 가중치 책정 방법이 중요하다. 하지만 SMS 문자 특성상 문장들이 짧기 때문에 출현하는 자질의 수가 적어 분류의 어려움을 겪게 된다. 이 같은 문제를 개선하기 위하여 본 논문에서는 슬라이딩 윈도우 기반 N-gram 확장을 통해 자질을 확장하고, 확장된 자질로 그래프를 구축하여 얕은 구조적 특징을 표현한다. 학습 데이터에 출현한 N-gram 자질을 정점(Vertex)으로, 자질의 출현 빈도를 그래프의 간선(Edge)의 가중치로 설정하여 햄(HAM)과 스팸(SPAM) 그래프를 각각 구성한다. 이렇게 구성된 그래프를 바탕으로 노드의 중요도와 간선의 가중치를 활용하여 최종적인 자질의 가중치를 결정한다. 입력 문자가 도착하면 스팸과 햄의 그래프를 각각 이용하여 입력 문자의 2개의 자질 벡터(Vector)를 생성한다. 생성된 자질 벡터를 지지 벡터 기계(Support Vector Machine)를 이용하여 각 SVM 확률 값(Probability Score)을 얻어 스팸 여부를 결정한다. 3가지의 실험환경에서 바이그램 자질과 이진 가중치를 사용한 기본 시스템보다 F1-Score의 약 최대 2.7%, 최소 0.5%까지 향상되었으며, 결과적으로 평균 약 1.35%의 성능 향상을 얻을 수 있었다.

• The KIPS Transactions:PartB
• /
• v.17B no.3
• /
• pp.249-254
• /
• 2010

We propose an automatic spam filter for e-mail data using Support Vector Machines(SVM). We use a lexical form of a word and its part of speech(POS) tags as features and select features by chi square statistics. We represent each feature by TF(text frequency), TF-IDF, and binary weight for experiments. After training SVM with the selected features, SVM classifies each e-mail as spam or not. In experiment, the selected features improve the performance of our system and we acquired overall 98.9% of accuracy with TREC05-p1 spam corpus.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.4
• /
• pp.35-41
• /
• 2001

This paper describes Linear Discriminant Analysis and common vector extraction for speech recognition. Voice signal contains psychological and physiological properties of the speaker as well as dialect differences, acoustical environment effects, and phase differences. For these reasons, the same word spelled out by different speakers can be very different heard. This property of speech signal make it very difficult to extract common properties in the same speech class (word or phoneme). Linear algebra method like BT (Karhunen-Loeve Transformation) is generally used for common properties extraction In the speech signals, but common vector extraction which is suggested by M. Bilginer et at. is used in this paper. The method of M. Bilginer et al. extracts the optimized common vector from the speech signals used for training. And it has 100％ recognition accuracy in the trained data which is used for common vector extraction. In spite of these characteristics, the method has some drawback-we cannot use numbers of speech signal for training and the discriminant information among common vectors is not defined. This paper suggests advanced method which can reduce error rate by maximizing the discriminant information among common vectors. And novel method to normalize the size of common vector also added. The result shows improved performance of algorithm and better recognition accuracy of 2％ than conventional method.

• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1313-1318
• /
• 2017

Aggregating local features in a single vector is a fundamental problem in an image search. In this process, the image search process can be speeded up if binary features which are extracted almost two order of magnitude faster than gradient-based features are utilized. However, in order to utilize the binary features in an image search, it is necessary to study the techniques for clustering binary features to generate binary visual words. This investigation is necessary because traditional clustering techniques for gradient-based features are not compatible with binary features. To this end, this paper studies the techniques for clustering binary features for the purpose of generating binary visual words. Through experiments, we analyze the trade-off between the accuracy and computational efficiency of an image search using binary features, and we then compare the proposed techniques. This research is expected to be applied to mobile applications, real-time applications, and web scale applications that require a fast image search.

• Journal of KIISE
• /
• v.43 no.8
• /
• pp.902-909
• /
• 2016

Question answering system (QA system) is a system that finds an actual answer to the question posed by a user, whereas a typical search engine would only find the links to the relevant documents. Recent works related to the open domain QA systems are receiving much attention in the fields of natural language processing, artificial intelligence, and data mining. However, the prior works on QA systems simply replace all words that are not in the training data with a single token, even though such unseen words are likely to play crucial roles in differentiating the candidate answers from the actual answers. In this paper, we propose a method to compute vectors of such unseen words by taking into account the context in which the words have occurred. Next, we also propose a model which utilizes inverse document frequencies (IDF) to efficiently process unseen words by expanding the system's vocabulary. Finally, we validate that the proposed method and model improve the performance of a QA system through experiments.

• Journal of KIISE:Computing Practices and Letters
• /
• v.15 no.4
• /
• pp.299-303
• /
• 2009

Patent information search is used for checking existence of earlier works. In patent information search, there are many reasons that fails to get appropriate information. This research proposes a method extracting alternative word candidates in order to minimize search failure due to keyword mismatch. Assuming that two words have similar meaning if they have similar co-occurrence words, the proposed method uses the concept of concentration, association word set, cosine similarity between association word sets and a ranking modification technique. Performance of the proposed method is evaluated using a manually extracted alternative word candidate list. Evaluation results show that the proposed method outperforms the document vector space model in recall.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.2
• /
• pp.96-100
• /
• 2006

Kim et al. Proposed Normalized Confidence Measure (NCM) [1-2] and it was successfully used for rejecting mis-recognized words in isolated word recognition. However their experiments were performed on the fixed word speech recognition. In this Paper we apply NCM to the domain of vocabulary independent speech recognition (VISP) and shows the rejection Performance of NCM in VISP. Specialty we Propose vector quantization (VQ) based method for overcoming the problem of unseen triphones. It is because NCM uses the statistics of triphone confidence in the case of triphone-based normalization. According to speech recognition experiments Phone-based normalization method shows better results than RLJC[3] and also triphone-based normalization approach. This results are different with those of Kim et al [1-2]. Concludingly the Phone-based normalization shows robust Performance in VISP domain.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.12 no.2
• /
• pp.115-122
• /
• 2002

This paper proposes a robust speech recognition algorithm based on the wavelet filter banks. Since the proposed algorithm adopts a multiple band decision-making scheme, it performs robustness for noise as the presence of noisy severely degrades the performance of speech recognition system. For evaluating the performance of the proposed scheme, we compared it with the conventional speech recognizer based on the VQ for the 10-isolated korean digits with car noise. Here, the proposed method showed more 9~27% improvement of the recognition rate than the conventional VQ algorithm for the various car noisy environments.

• The KIPS Transactions:PartB
• /
• v.12B no.3 s.99
• /
• pp.357-364
• /
• 2005

In this paper, we propose an accurate and fast keyword spotting system for searching user-specified keyword in Hangul document images by using two-level image-to-image matching. The system is composed of character segmentation, creating a query image, feature extraction, and matching procedure. Two different feature vectors are used in the matching procedure. An experiment using 1600 Hangul word images from 8 document images, downloaded from the website of Korea Information Science Society, demonstrates that the proposed system is superior to conventional image-based document retrieval systems.

• Journal of the Institute of Convergence Signal Processing
• /
• v.1 no.1
• /
• pp.32-41
• /
• 2000

In this paper, we proposed a method which predict the verb by using the neural network in order to enhance communication rate in augmentative communication system for Korean language disorders. Each word is represented by an information vector according to syntax and semantics, and is positioned at the state space by being partitioned into various regions different from a dictionary-like lexicon. Conceptual similarity is realized through position in state space. When a symbol was pressed, we could find the word for the symbol at the position in the state space. In order to prevent verb prediction's redundancy according to input units, we predicted the verb after separating class using the neural network. In the result we can enhance $20\% communication rate in the restricted space