Digital Library[ Search Result ]
SCA: Improving Document Grounded Response Generation based on Supervised Cross-Attention
Hyeongjun Choi, Seung-Hoon Na, Beomseok Hong, Youngsub Han, Byoung-Ki Jeon
http://doi.org/10.5626/JOK.2024.51.4.326
Document-grounded response generation is the task of aiming at generating conversational responses by “grounding” the factual evidence on task-specific domain, such as consumer consultation or insurance planning, where the evidence is obtained from the retrieved relevant documents in response to a user’s question under the current dialogue context. In this study, we propose supervised cross-attention (SCA) to enhance the ability of the response generation model to find and incorporate “response-salient snippets” (i.e., spans or contents), which are parts of the retrieved document that should be included and maintained in the actual answer generation. SCA utilizes the additional supervised loss that focuses cross-attention weights on the response-salient snippets, and this attention supervision likely enables a decoder to effectively generate a response in a “saliency-grounding” manner, by strongly attending to the important parts in the retrieved document. Experiment results on MultiDoc2Dial show that the use of SCA and additional performance improvement methods leads to the increase of 1.13 in F1 metric over the existing SOTA, and reveals that SCA leads to the increase of 0.25 in F1.
Defect Detection of Bridge based on Impact-Echo Signals and Long Short-Term Memory
Byoung-Doo Oh, Hyung Choi, Young Jin Kim, Won Jong Chin, Yu-Seop Kim
http://doi.org/10.5626/JOK.2021.48.9.988
PSC box girder bridges comprise various elements, of which the tendon is the most important. The tendon is exceedingly vulnerable to defects such as corrosion. So, to protect it, it is inserted into the duct and then poured concrete. However, because the inner diameter of the duct is exceedingly narrow, there is a high probability that a cavity can occur with even a minor mistake. Thus, defects are identified through the professional interpretation of signals obtained using non-destructive testing. However, this requires high cost and much time. Additionally, it is difficult to accurately identify the internal state of the concrete structure. Thus, this study intends to apply Long Short-Term Memory. In this case, the Impact-Echo method, which is often used for concrete structures, is used as an input feature. And the characteristics of the structure (concrete thickness, depth of duct, distance between the hitting point and the measuring point) are used for learning. And, the frequency component of the IE signals is additionally used. As a result, the proposed model in this study can confirm approximately 88.56% accuracy.
Design and Implementation of an In-Memory File System Cache with Selective Compression
http://doi.org/10.5626/JOK.2017.44.7.658
The demand for large-scale storage systems has continued to grow due to the emergence of multimedia, social-network, and big-data services. In order to improve the response time and reduce the load of such large-scale storage systems, DRAM-based in-memory cache systems are becoming popular. However, the high cost of DRAM severely restricts their capacity. While the method of compressing cache entries has been proposed to deal with the capacity limitation issue, compression and decompression, which are technically difficult to parallelize, induce significant processing overhead and in turn retard the response time. A selective compression scheme is proposed in this paper for in-memory file system caches that rapidly estimates the compression ratio of incoming cache entries with their Shannon entropies and compresses cache entries with low compression ratio. In addition, a description is provided of the design and implementation of an in-kernel in-memory file system cache with the proposed selective compression scheme. The evaluation showed that the proposed scheme reduced the execution time of benchmarks by approximately 18% in comparison to the conventional non-compressing in-memory cache scheme. It also provided a cache hit ratio similar to the all-compressing counterpart and reduced 7.5% of the execution time by reducing the compression overhead. In addition, it was shown that the selective compression scheme can reduce the CPU time used for compression by 28% compared to the case of the all-compressing scheme.
Optimizing Constant Value Generation in Just-in-time Compiler for 64-bit JavaScript Engine
JavaScript is widely used in web pages with HTML. Many JavaScript engines adopt Just-in-time compilers to accelerate the execution of JavaScript programs. Recently, many newly introduced devices are adopting 64-bit CPUs instead of 32-bit and Just-in-time compilers for 64-bit CPU are slowly being introduced in JavaScript engines. However, there are many inefficiencies in the currently available Just-in-time compilers for 64-bit devices. Especially, the size of code is significantly increased compared to 32-bit devices, mainly due to 64-bit wide addresses in 64-bit devices. In this paper, we are going to address the inefficiencies introduced by 64-bit wide addresses and values in the Just-in-time compiler for the V8 JavaScript engine and propose more efficient ways of generating constant values and addresses to reduce the size of code. We implemented the proposed optimization in the V8 JavaScript engine and measured the size of code as well as performance improvements with Octane and SunSpider benchmarks. We observed a 3.6% performance gain and 0.7% code size reduction in Octane and a 0.32% performance gain and 2.8% code size reduction in SunSpider.
Study on the Improvement about User Authentication of Android Third Party Application Through the Vulnerability in Google Voice
Seyeong Lee, Jaekyun Park, Sungdae Hong, Hyoungki Choi
In the Android market, a large portion of the market share consists of third party applications, but not much research has been performed in this respect. Of these applications, mobile Voice Over IP (VoIP) applications are one of the types of applications that are used the most. In this paper, we focus on user authentication methods for three representative applications of the Google Voice service, which is a famous mobile VoIP application. Then, with respect to the Android file system, we developed a method to store and to send user information for authentication. Finally, we demonstrate a vulnerability in the mechanism and propose an improved mechanism for user authentication by using hash chaining and an elliptic curve Diffie-Hellman key exchange.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr