• Journal of Korea Multimedia Society
• /
• v.11 no.11
• /
• pp.1471-1479
• /
• 2008

In this paper, we propose a technique to improve Out-Of-Vocabulary(OOV) rejection algorithms in variable vocabulary recognition system which is much used in ASR(Automatic Speech Recognition). The rejection system can be classified into two categories by their implementation method, keyword spotting method and utterance verification method. The utterance verification method uses the likelihood ratio of each phoneme Viterbi score relative to anti-phoneme score for deciding OOV. In this paper, we add speaker verification system before utterance verification and calculate an speaker verification probability. The obtained speaker verification probability is applied for determining the proposed variable-confidence threshold. Using the proposed method, we achieve the significant performance improvement; CA(Correctly Accepted for keyword) 94.23%, CR(Correctly Rejected for out-of-vocabulary) 95.11% in office environment, and CA 91.14%, CR 92.74% in noisy environment.

• 한국데이터정보과학회:학술대회논문집
• /
• 2005.10a
• /
• pp.177-188
• /
• 2005

Association rule mining searches for interesting relationships among items in a given large data set. Association rules are frequently used by retail stores to assist in marketing, advertising, floor placement, and inventory control. There are three primary threshold measures in association rule; support and confidence and lift. In the case of appling real world to association rules, we have some difficulties in data interpretation because we obtain many rules. In this paper, we develop the model of association rules using latent variables for environmental survey data.

• Journal of the Korean Data and Information Science Society
• /
• v.17 no.1
• /
• pp.149-160
• /
• 2006

Association rule mining searches for interesting relationships among items in a given large data set. Association rules are frequently used by retail stores to assist in marketing, advertising, floor placement, and inventory control. There are three primary threshold measures in association rule; support and confidence and lift. In the case of appling real world to association rules, we have some difficulties in data interpretation because we obtain many rules. In this paper, we develop the model of association rules using latent variables for environmental survey data.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.5
• /
• pp.891-899
• /
• 2010

Association rule is the technique to represent the relationship between two or more items by numerical representing for the relevance of each item in vast amounts of databases, and is most being used in data mining. The basic thresholds for association rule are support, confidence, and lift. these are used to generate the association rules. We need standardization of lift because the range of lift value is different from that of support and confidence. And also we need standardization of support and confidence to compare objectively association level of antecedent variables for one descendant variable. In this paper we propose a method for standardization of association thresholds considering marginal probability for each item to grasp objectively and exactly association level, check the conditions for association criteria and then compare association thresholds with standardized association thresholds using some concrete examples.

• Journal of Trauma and Injury
• /
• v.35 no.1
• /
• pp.25-33
• /
• 2022

Purpose: The scoring system for traumatic liver injury (SSTLI) was developed in 2015 to predict mortality in patients with polytraumatic liver injury. This study aimed to validate the SSTLI as a prognostic factor in patients with polytrauma and liver injury through a generalized estimating equation analysis. Methods: The medical records of 521 patients with traumatic liver injury from January 2015 to December 2019 were reviewed. The primary outcome variable was in-hospital mortality. All the risk factors were analyzed using multivariate logistic regression analysis. The SSTLI has five clinical measures (age, Injury Severity Score, serum total bilirubin level, prothrombin time, and creatinine level) chosen based on their predictive power. Each measure is scored as 0-1 (age and Injury Severity Score) or 0-3 (serum total bilirubin level, prothrombin time, and creatinine level). The SSTLI score corresponds to the total points for each item (0-11 points). Results: The areas under the curve of the SSTLI to predict mortality on post-traumatic days 0, 1, 3, and 5 were 0.736, 0.783, 0.830, and 0.824, respectively. A very good to excellent positive correlation was observed between the probability of mortality and the SSTLI score (γ=0.997, P<0.001). A value of 5 points was used as the threshold to distinguish low-risk (<5) from high-risk (≥5) patients. Multivariate analysis using the generalized estimating equation in the logistic regression model indicated that the SSTLI score was an independent predictor of mortality (odds ratio, 1.027; 95% confidence interval, 1.018-1.036; P<0.001). Conclusions: The SSTLI was verified to predict mortality in patients with polytrauma and liver injury. A score of ≥5 on the SSTLI indicated a high-risk of post-traumatic mortality.