• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.4
• /
• pp.145-152
• /
• 2020

MRC (Machine reading comprehension) is the AI NLP task that predict the answer for user's query by understanding of the relevant document and which can be used in automated consult services such as chatbots. Recently, the BERT (Pre-training of Deep Bidirectional Transformers for Language Understanding) model, which shows high performance in various fields of natural language processing, have two phases. First phase is Pre-training the big data of each domain. And second phase is fine-tuning the model for solving each NLP tasks as a prediction. In this paper, we have made the Patent MRC dataset and shown that how to build the patent consultation training data for MRC task. And we propose the method to improve the performance of the MRC task using the Pre-trained Patent-BERT model by the patent consultation corpus and the language processing algorithm suitable for the machine learning of the patent counseling data. As a result of experiment, we show that the performance of the method proposed in this paper is improved to answer the patent counseling query.

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.353-356
• /
• 2007

With advances in sensor technology, current researches on the pertinent techniques are actively directed toward the way which enables the USN computing service. For many applications using sensor networks, the incoming data are by nature characterized as high-speed, continuous, real-time and infinite. Due to such uniqueness of sensor data characteristics, for some instances a finite-sized buffer may not accommodate the entire incoming data, which leads to inevitable loss of data, and requirement for fast processing makes it impossible to conduct a thorough investigation of data. In addition to the potential problem of loss of data, incoming data in its raw form may exhibit high degree of complexity which evades simple query or alerting services for capturing and extracting useful information. Furthermore, as traditional mining techniques are developed to handle fixed, static historical data, they are not useful and directly applicable for analyzing the sensor data. In this paper, (1) describe how three mining techniques (sensor data outlier analysis, sensor pattern analysis, and sensor data prediction analysis) are appropriate for the USN middleware structure, with their application to the stream data in ocean environment. (2) Another proposal is a middleware structure based on USN environment adaptive to above mining techniques. This middleware structure includes sensor nodes, sensor network common interface, sensor data processor, sensor query processor, database, sensor data mining engine, user interface and so on.

• Journal of Korean Society for Geospatial Information Science
• /
• v.19 no.3
• /
• pp.13-21
• /
• 2011

In mobile environment supporting mobility technologies, user requirements have been increased with respect to utilization of location information. In particular, moving object DBMS has consistently posed in order to efficiently maintain traffic information related to location of vehicle which tents to tremendously change over time. Despite the fact that these sorts of researches must be taken into consideration, empirical studies on moving object in terms of map database for lbs service, spatial attribute of which is continuously changed over time, have rarely performed. Therefore, aim of this paper is to suggest efficient spatial index scheme, which is capable of supporting query processing algorithm and location of moving object over time, by developing new empirical model. As a result, we can come to the conclusion that moving object model based on multi-fixed grid index makes it possible to cut down on the number of entity for retrieving. What's more, this model enables hierarchical data to be accessed through efficient spatial filtering on large-scale lbs data and constraints in accordance with level in order to display map.

• Journal of KIISE:Computing Practices and Letters
• /
• v.12 no.4
• /
• pp.219-228
• /
• 2006

Many biomedical research groups have been trying to share their outputs to increase the efficiency of research. As part of their efforts, a common ontology named Gene Ontology(GO), which comprises controlled vocabulary for the functions of genes, was built. However, data from many research groups are distributed and most systems don't support integrated semantic queries on them. Furthermore, the semantics of the associations between concepts from external classification systems and GO are still not clarified, which makes integrated semantic query infeasible. In this paper we present an ontology matching and integration system, called AutoGOA, which first resolves the semantics of the associations between concepts semi-automatically, and then constructs integrated ontology containing concepts from GO and external classification systems. Also we describe a web-based application, named GOGuide II, which allows the user to browse, query and visualize integrated data.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.26 no.4
• /
• pp.301-307
• /
• 2016

This paper proposes a lyric search method to consider the characteristics of the user query. According to the fact that queries for the lyric search are derived from highlight parts of the music, this paper uses the hierarchical agglomerative clustering to find the highlight and proposes a Gaussian weighting to consider the neighbor of the highlight as well as highlight. By setting the mean of a Gaussian weighting at the highlight, this weighting function has higher weights near the highlight and the lower weights far from the highlight. Then, this paper constructs a index of lyrics with the gaussian weighting. According to the experimental results on a data set obtained from 5 real users, the proposed method is proved to be effective.

• The KIPS Transactions:PartD
• /
• v.8D no.5
• /
• pp.605-612
• /
• 2001

In order to solve the overload of academic affairs management system, we propose a method builds a distributed Replication server and uses this server with the present centralized system. Normal query transactions which are not required for data modification are composed of almost all DML sentences. So we construct the distributed replication servers according to the data characteristics and make them perform the query transaction without modification. In this way, we can simultaneously distribute users and data, and cut down processing time for every transaction. Also Replication server has the advantages of implemental efficiency and economical because it uses resources of present centralized system without and additional configurations. Usually, to distribute the overload of server, they can use way, Client-side overload distribution that user program get present overload status then can choose a less overloaded server, and the other way, Server-side overload distribution that make use of Application Layer Scheduling Technique and IP Layer Scheduling Technique. Our Replication server can reduce the overload of centralized system by eliminating or complementing those defects of overload distribution, referred to in the forehead.

• The KIPS Transactions:PartD
• /
• v.17D no.1
• /
• pp.17-24
• /
• 2010

Recently, there has been growing interest in skyline queries. Most of works for skyline queries assume that the data do not have null value. However, when we input data through the Web or with other different tools, there exist incomplete data with null values. As a result, several skyline processing techniques for incomplete data have been proposed. However, available skyline query techniques for incomplete data do not consider the environments that coexist complete data and incomplete data since these techniques deal with the incomplete data only. In this paper, we propose a novel skyline group processing technique which evaluates skyline queries for the environments that coexist complete data and incomplete data. To do this, we introduce the top-k(g) skyline group query which searches g skyline groups with respect to the user's dimensional preference. In our experimental study, we show efficiency of our proposed technique.

• Journal of KIISE:Computing Practices and Letters
• /
• v.5 no.5
• /
• pp.628-638
• /
• 1999

In this paper, we presents a method of retrieving 24 bpp RGB images based on color-spatial features. For each image, it is subdivided into regions by using similarity of color after converting RGB color space to CIE L*u*v* color space that is perceptually uniform. Our segmentation algorithm constrains the size of region because a small region is discardable and a large region is difficult to extract spatial feature. For each region, averaging color and center of region are extracted to construct color-spatial features. During the image retrieval process, the color and spatial features of query are compared with those of the database images using our similarity measure to determine the set of candidate images to be retrieved. We implement a content-based color image retrieval system using the proposed method. The system is able to retrieve images by user graphic or example image query. Experimental results show that Recall/Precision is 0.80/0.84.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.7
• /
• pp.877-883
• /
• 2021

Artificial intelligence technology is widely used in industrial and smart home fields such as manufacturing robots, artificial intelligence speakers, and robot vacuum cleaners. In this paper, we designed and implemented a 1:1 chatbot system based on artificial intelligence for use in RCMS (Real-time Cash Management System). The RCMS chatbot implemented in this paper was constructed with a total of 210 query scenarios in nine areas, including research expenses and system usage, based on 13,500 questions and answers from existing online bulletin boards. The chatbot is expected to solve the problem of insufficient number of counselors and to increase user satisfaction by responding to the researcher's inquiries after working hours, and the recommendation service for the cost of use, which had the most inquiries from researchers, reduces the number of consultations. It is expected to improve the quality of answers to other counseling inquiries.

• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.4
• /
• pp.91-103
• /
• 2024

A tourist destination ranking system was designed that employs a semantic search to extract information with reasonable accuracy. To this end the process involves collecting data, preprocessing text reviews of tourist spots, and embedding the corpus and queries with SBERT. We calculate the similarity between data points, filter out those below a specified threshold, and then rank the remaining tourist destinations using a count-based algorithm to align them semantically with the query. To assess the efficacy of the ranking algorithm experiments were conducted with four queries. Furthermore, 58,175 sentences were directly labeled to ascertain their semantic relevance to the third query, 'crowdedness'. Notably, human-labeled data for crowdedness showed similar results. Despite challenges including optimizing thresholds and imbalanced data, this study shows that a semantic search is a powerful method for understanding user intent and recommending tourist destinations with less time and costs.