• Title/Summary/Keyword: Multi-Label

Search Result 220, Processing Time 0.022 seconds

Multi-decoder Model Reflecting Candidate Label Information (후보 레이블 정보를 반영한 멀티 디코더 모델)

  • Park, Won-Jae;Choi, Gi-Hyeon;Kim, Hark-Soo
    • Annual Conference on Human and Language Technology
    • /
    • 2021.10a
    • /
    • pp.307-310
    • /
    • 2021
  • 지도 학습을 하기 위해선 레이블이 부착된 데이터셋이 필요하다. 크라우드소싱 서비스를 통해 데이터셋을 구축하는데 다수의 주석자(Annotator)가 관여한다. 다수의 주석자가 레이블을 할당하고 과반수인 레이블을 최종 정답으로 결정한다. 이 과정에서 최종 정답과 다른 후보 레이블의 정보가 누락된다. 이를 완화하고 목표 작업에 대한 성능을 높이기 위해 후보 레이블에 대한 정보를 반영하는 멀티 디코더 모델을 제안한다. KLUE-TC, SNLI, MNLI 데이터셋으로 정량적 성능 평가를 수행하였으며 실험한 데이터셋 모두 일괄적인 성능 향상을 보였다.

  • PDF

Web Application for Creating Emotional ID Photos using Deep Learning (딥러닝을 활용한 감성 증명사진 제작 웹 애플리케이션)

  • Kim, Do Young;Kang, In Yeong;Kim, Yeon Su;Park, Goo man
    • Proceedings of the Korean Society of Broadcast Engineers Conference
    • /
    • 2022.06a
    • /
    • pp.1261-1264
    • /
    • 2022
  • 최근 본인에게 어울리는 색상을 배경으로 촬영하는 감성 증명사진이 유행하고 있다. 개인마다 퍼스널 컬러를 찾아 배경색에 적용하는 것은 시간, 비용, 인력적으로 어려움이 있으므로 자동으로 개인에 따른 배경색을 찾아서 사진을 합성하여 감성 증명사진을 제작해 주는 딥러닝 기반 시스템을 구축하였다. 본 논문에서는 Convolution Neural Network 를 기반으로 한 딥러닝 기술을 이용해 Image Matting 과 Multi-Label Classification 을 수행하여 기존 감성 증명사진들을 학습하여 모델을 구축하였으며, 해당 시스템으로 사용자에게 새로운 배경색이 적용된 감성 증명사진을 제공하는 웹 애플리케이션을 제안한다.

  • PDF

STag: Supernova Tagging and Classification

  • Davison, William;Parkinson, David;Tucker, Brad E.
    • The Bulletin of The Korean Astronomical Society
    • /
    • v.46 no.2
    • /
    • pp.45.3-46
    • /
    • 2021
  • Supernovae classes have been defined phenomenologically, based on spectral features and time series data, since the specific details of the physics of the different explosions remain unrevealed. However, the number of these classes is increasing as objects with new features are observed, and the next generation of large-surveys will only bring more variety to our attention. We apply the machine learning technique of multi-label classification to the spectra of supernovae. By measuring the probabilities of specific features or 'tags' in the supernova spectra, we can compress the information from a specific object down to that suitable for a human or database scan, without the need to directly assign to a reductive 'class'. We use logistic regression to assign tag probabilities, and then a feed-forward neural network to filter the objects into the standard set of classes, based solely on the tag probabilities. We present STag, a software package that can compute these tag probabilities and make spectral classifications.

  • PDF

Multi-label Open Intent Classification using Known Intent Information (의도 정보를 활용한 다중 레이블 오픈 의도 분류)

  • Nahyeon Park;Seongmin Cho;Hyun-Je Song
    • Annual Conference on Human and Language Technology
    • /
    • 2023.10a
    • /
    • pp.479-484
    • /
    • 2023
  • 다중 레이블 오픈 의도 분류란 다중 의도 분류와 오픈 의도 분류가 합쳐져 오픈 도메인을 가정하고 진행하는 다중 의도 분류 문제이다. 발화 속에는 여러 의도들이 존재한다. 이때 사전에 정의된 의도 여부만을 판별하는 것이 아니라 사전에 정의되어 있는 의도에 대해서만이라도 어떤 의도인지 분류할 수 있어야 한다. 본 논문에서는 발화 속 의도 정보를 활용하여 다중 레이블 오픈 의도를 분류하는 모델을 제안한다. 먼저, 문장의 의도 개수를 예측한다. 그리고 다중 레이블 의도 분류기를 통해 다중 레이블 의도 분류를 진행하여 의도 정보를 획득한다. 획득한 의도 정보 속 다중 의도 개수와 전체 의도 개수를 비교하여 전체 의도 개수가 더 많다면 오픈 의도가 존재한다고 판단한다. 실험 결과 제안한 방법은 MixATIS의 75% 의도에서 정확도 94.49, F1 97.44, MixSNIPS에서는 정확도 86.92, F1 92.96의 성능을 보여준다.

  • PDF

Smart Contract Vulnerability Detection Study Based on Control Flow Graphs (제어 흐름 그래프 기반 스마트 컨트랙트 취약성 탐지 연구)

  • Yoo-Young Cheong;La Yeon Choi;Dong-Hyuk Im
    • Proceedings of the Korea Information Processing Society Conference
    • /
    • 2023.11a
    • /
    • pp.1247-1249
    • /
    • 2023
  • 스마트 컨트랙트는 블록체인 상에서 실행되는 프로그램으로 복잡한 비즈니스 논리를 처리할 수 있다. 그러나 블록체인의 무결성과 조건에 따라 실행되는 특성을 이용한 악의적 사용으로 인하여 블록체인 보안에서 시급한 문제가 되고있다. 따라서 스마트 컨트랙트 취약성 탐지문제는 최근 많은 연구가 이루어지고 있다. 그러나 기존 연구의 대부분이 단일 유형의 취약성 여부에 대한 탐지에만 초점이 맞춰져 있어 여러 유형의 취약성에 대한 동시 식별이 어렵다. 이 문제를 해결하고자 본 연구에서는 스마트 컨트랙트 소스코드 제어 흐름 그래프를 기반으로 그래프의 forward edge와 backward edge를 고려한 신경망으로 그래프 구조를 학습한 후 그래프 multi-label classification을 진행하여 다중 취약성을 탐지할 수 있는 모델을 제안한다.

Semantic Feature Analysis for Multi-Label Text Classification on Topics of the Al-Quran Verses

  • Gugun Mediamer;Adiwijaya
    • Journal of Information Processing Systems
    • /
    • v.20 no.1
    • /
    • pp.1-12
    • /
    • 2024
  • Nowadays, Islamic content is widely used in research, including Hadith and the Al-Quran. Both are mostly used in the field of natural language processing, especially in text classification research. One of the difficulties in learning the Al-Quran is ambiguity, while the Al-Quran is used as the main source of Islamic law and the life guidance of a Muslim in the world. This research was proposed to relieve people in learning the Al-Quran. We proposed a word embedding feature-based on Tensor Space Model as feature extraction, which is used to reduce the ambiguity. Based on the experiment results and the analysis, we prove that the proposed method yields the best performance with the Hamming loss 0.10317.

The Effect of Brand Extension of Private Label on Consumer Attitude - a focus on the moderating effect of the perceived fit difference between parent brands and an extended brand - (PL의 브랜드확장이 소비자태도에 미치는 영향에 관한 연구 : 모브랜드 적합도 인식 차이의 조절효과를 중심으로)

  • Kim, Jong-Keun;Kim, Hyang-Mi;Lee, Jong-Ho
    • Journal of Distribution Research
    • /
    • v.16 no.4
    • /
    • pp.1-27
    • /
    • 2011
  • Introduction: Sales of private labels(PU have been growing m recent years. Globally, PLs have already achieved 20% share, although between 25 and 50% share in most of the European markets(AC. Nielson, 2005). These products are aimed to have comparable quality and prices as national brand(NB) products and have been continuously eroding manufacturer's national brand market share. Stores have also started introducing premium PLs that are of higher-quality and more reasonably priced compared to NBs. Worldwide, many retailers already have a multiple-tier private label architecture. Consumers as a consequence are now able to have a more diverse brand choice in store than ever before. Since premium PLs are priced higher than regular PLs and even, in some cases, above NBs, stores can expect to generate higher profits. Brand extensions and private label have been extensively studied in the marketing field. However, less attention has been paid to the private label extension. Therefore, this research focuses on private label extension using the Multi-Attribute Attitude Model(Fishbein and Ajzen, 1975). Especially there are few studies that consider the hierarchical effect of the PL's two parent brands: store brand and the original PL. We assume that the attitude toward each of the two parent brands affects the attitude towards the extended PL. The influence from each parent brand toward extended PL will vary according to the perceived fit between each parent brand and the extended PL. This research focuses on how these two parent brands act as reference points to one another in the consumers' choice consideration. Specifically we seek to understand how store image and attitude towards original PL affect consumer perceptions of extended premium PL. How consumers perceive extended premium PLs could provide strategic suggestions for retailer managers with specific suggestions on whether it is more effective: to position extended premium PL similarly or dissimilarly to original PL especially on the quality dimension and congruency with store image. There is an extensive body of research on branding and brand extensions (e.g. Aaker and Keller, 1990) and more recently on PLs(e.g. Kumar and Steenkamp, 2007). However there are no studies to date that look at the upgrading and influence of original PLs and attitude towards store on the premium PL extension. This research wishes to make a contribution to this gap using the perceived fit difference between parent brands and extended premium PL as the context. In order to meet the above objectives, we investigate which factors heighten consumers' positive attitude toward premium PL extension. Research Model and Hypotheses: When considering the attitude towards the premium PL extension, we expect four factors to have an influence: attitude towards store; attitude towards original PL; perceived congruity between the store image and the premium PL; perceived similarity between the original PL and the premium PL. We expect that all these factors have an influence on consumer attitude towards premium PL extension. Figure 1 gives the research model and hypotheses. Method: Data were collected by an intercept survey conducted on consumers at discount stores. 403 survey responses were attained (total 59.8% female, across all age ranges). Respondents were asked to respond to a series of Questions measured on 7 point likert-type scales. The survey consisted of Questions that measured: the trust towards store and the original PL; the satisfaction towards store and the original PL; the attitudes towards store, the original PL, and the extended premium PL; the perceived similarity of the original PL and the extended premium PL; the perceived congruity between the store image and the extended premium PL. Product images with specific explanations of the features of premium PL, regular PL and NB we reused as the stimuli for the Question response. We developed scales to measure the research constructs. Cronbach's alphaw as measured each construct with the reliability for all constructs exceeding the .70 standard(Nunnally, 1978). Results: To test the hypotheses, path analysis was conducted using LISREL 8.30. The path analysis for verification of the model produced satisfactory results. The validity index shows acceptable results(${\chi}^2=427.00$(P=0.00), GFI= .90, AGFI= .87, NFI= .91, RMSEA= .062, RMR= .047). With the increasing retailer use of premium PLBs, the intention of this research was to examine how consumers use original PL and store image as reference points as to the attitude towards premium PL extension. Results(see table 1 & 2) show that the attitude of each parent brand (attitudes toward store and original pL) influences the attitude towards extended PL and their perceived fit moderates these influences. Attitude toward the extended PL was influenced by the relative level of perceived fit. Discussion of results and future direction: These results suggest that the future strategy for the PL extension needs to consider that positive parent brand attitude is more strongly associated with the attitude toward PL extensions. Specifically, to improve attitude towards PL extension, building and maintaining positive attitude towards original PL is necessary. Positioning premium PL congruently to store image is also important for positive attitude. In order to improve this research, the following alternatives should also be considered. To improve the research model's predictive power, more diverse products should be included in study. Other attributes of product should also be included such as design, brand name since we only considered trust and satisfaction as factors to build consumer attitudes.

  • PDF

Hierarchical Neural Network for Real-time Medicine-bottle Classification (실시간 약통 분류를 위한 계층적 신경회로망)

  • Kim, Jung-Joon;Kim, Tae-Hun;Ryu, Gang-Soo;Lee, Dae-Sik;Lee, Jong-Hak;Park, Kil-Houm
    • Journal of the Korean Institute of Intelligent Systems
    • /
    • v.23 no.3
    • /
    • pp.226-231
    • /
    • 2013
  • In The matching algorithm for automatic packaging of drugs is essential to determine whether the canister can exactly refill the suitable medicine. In this paper, we propose a hierarchical neural network with the upper and lower layers which can perform real-time processing and classification of many types of medicine bottles to prevent accidental medicine disaster. A few number of low-dimensional feature vector are extracted from the label images presenting medicine-bottle information. By using the extracted feature vectors, the lower layer of MLP(Multi-layer Perceptron) neural networks is learned. Then, the output of the learned middle layer of the MLP is used as the input to the upper layer of the MLP learning. The proposed hierarchical neural network shows good classification performance and real- time operation in the test of up to 30 degrees rotated to the left and right images of 100 different medicine bottles.

A Text Sentiment Classification Method Based on LSTM-CNN

  • Wang, Guangxing;Shin, Seong-Yoon;Lee, Won Joo
    • Journal of the Korea Society of Computer and Information
    • /
    • v.24 no.12
    • /
    • pp.1-7
    • /
    • 2019
  • With the in-depth development of machine learning, the deep learning method has made great progress, especially with the Convolution Neural Network(CNN). Compared with traditional text sentiment classification methods, deep learning based CNNs have made great progress in text classification and processing of complex multi-label and multi-classification experiments. However, there are also problems with the neural network for text sentiment classification. In this paper, we propose a fusion model based on Long-Short Term Memory networks(LSTM) and CNN deep learning methods, and applied to multi-category news datasets, and achieved good results. Experiments show that the fusion model based on deep learning has greatly improved the precision and accuracy of text sentiment classification. This method will become an important way to optimize the model and improve the performance of the model.

Generating Multiple Paths by Using Multi-label Vine-building Shortest Path Algorithm (수정형 덩굴망 최단경로 탐색 알고리즘을 이용한 다경로 생성 알고리즘의 개발)

  • Kim, Ik-Ki
    • Journal of Korean Society of Transportation
    • /
    • v.22 no.2 s.73
    • /
    • pp.121-130
    • /
    • 2004
  • In these days, multiple-path generation method is highly demanded in practice and research areas, which can represents realistically travelers behavior in choosing possible alternative paths. The multiple-path generation algorithm is one of the key components for policy analysis related to ATIS, DRGS and ATMS in ITS. This study suggested a method to generate multiple Possible paths from an origin to a destination. The approach of the suggested method is different from an other existing methods(K-shortest path algorithm) such as link elimination approach, link penalty approach and simulation approach. The result of the multi-label vine-building shortest path algorithm(MVA) by Kim (1998) and Kim(2001) was used to generate multiple reasonable possible paths with the concept of the rational upper boundary. Because the MVA algorithm records the cost, back-node and back-back node of the minimum path from the origin to the concerned node(intersection) for each direction to the node, many potential possible paths can be generated by tracing back. Among such large number of the potential possible paths, the algorithm distinguishes reasonable alternative paths from the unrealistic potential possible paths by using the concept of the rational upper boundary. The study also shows the very simple network examples to help the concept of the suggested path generation algorithm.