• Proceedings of the Korean Information Science Society Conference
• /
• 2010.06c
• /
• pp.349-352
• /
• 2010

질의를 이용한 정보 검색 기술에서 단어 의미의 모호성에 의해 사용자가 검색 하고자 하는 주제 이외의 문서 까지 검색되고 있다. 이러한 문제는 모바일기기의 검색 환경에서 두드러진다. 모바일에서의 검색은 문서의 로딩속도가 느리며 작은 화면에 의해 스크롤이 잦다. 그러므로 원하는 검색 결과가 검색 첫 페이지 이외에 위치하거나, 또는 페이지 하단에 위치할 경우 검색 결과를 확인하는 대에 많은 시간과 노력이 필요하다. 이러한 문제를 해결하기위해선 단어 의미의 모호성을 해결하고 사용자가 검색하고자하는 주제의 검색결과를 검색 상위에 위치시킬 수 있는 방법을 필요로 한다. 이 연구에서는 연관 단어 추출과 TF*IDF를 이용하여, 검색결과를 re-ranking하는 방법을 제시한다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.442-446
• /
• 2022

자동 음성 인식(Automatic Speech Recognition) 혹은 Speech-to-Text(STT)는 컴퓨터가 사람이 말하는 음성 언어를 텍스트 데이터로 전환하는 일련의 처리나 기술 등을 일컫는다. 음성 인식 기술이 다양한 산업 전반에 걸쳐 적용됨에 따라 높은 수준의 정확도와 더불어 다양한 분야에 적용할 수 있는 음성 인식 기술에 대한 필요성이 점차 증대되고 있다. 다만 한국어 음성 인식의 경우 기존 선행 연구에 비해 예사말/높임말의 구분이나 어미, 조사 등의 인식에 어려움이 있어 음성 인식 결과 후처리를 통한 성능 개선이 중요하다. 따라서 본 논문에서는 N-Best 음성 인식 결과가 구성되었을 때 Re-ranking을 통해 한국어 음성 인식의 성능을 개선하는 모델을 제안한다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.12 no.2
• /
• pp.69-76
• /
• 2012

This paper proposes a novel way of personalized web search through re-ranking the search results with user profiles of concept-network structure. Basically, personalized search systems need to be based on user profiles that contain users' search patterns, and they actively use the user profiles in order to expand initial queries or to re-rank the search results. The proposed method is a sort of a re-ranking personalized search method integrated with query expansion facility. The method identifies some documents which occur commonly among a set of different search results from the expanded queries, and re-ranks the search results by the degree of co-occurring. We show that the proposed method outperforms the conventional ones by performing the empirical web search with a number of actual users who have diverse information needs and query intents.

• Journal of the Korean Society for Library and Information Science
• /
• v.57 no.3
• /
• pp.249-277
• /
• 2023

The purpose of this study is to explore the potential contribution of citation metrics to improving the search performance of citation index databases. To this end, the study generated ten queries in the field of library and information science and conducted experiments based on the relevance assessment using 3,467 documents retrieved from the Web of Science and 60,734 documents published in 85 SSCI journals in the field of library and information science from 2000 to 2021. The experiments included re-ranking of the top 100 search results using citation metrics and search methods, query expansion experiments using vector space model retrieval systems, and the construction of a citation-based re-ranking system. The results are as follows: 1) Re-ranking using citation metrics differed from Web of Science's performance, acting as independent metrics. 2) Combining query term frequencies and citation counts positively affected performance. 3) Query expansion generally improved performance compared to the vector space model baseline. 4) User-based query expansion outperformed system-based. 5) Combining citation counts with suitability documents affected ranking within top suitability documents.

• Journal of KIISE:Software and Applications
• /
• v.31 no.11
• /
• pp.1519-1533
• /
• 2004

When Web-based special retrieval systems for scientific field extremely restrict the expression of user's information request, the process of the information content analysis and that of the information acquisition become inconsistent. In this paper, we apply the fuzzy retrieval model to solve the high time complexity of the retrieval system by constructing a reduced term set for the term's relatively importance degree. Furthermore, we perform a cluster retrieval to reflect the user's Query exactly through the similarity relation matrix satisfying the characteristics of the fuzzy compatibility relation. We have proven the performance of a proposed re-ranking model based on the similarity union of the fuzzy retrieval model and the document cluster retrieval model.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.1
• /
• pp.7-15
• /
• 2014

Image retrieval has become a huge part of computer vision and data mining. Although commercial image retrieval systems such as Google show great performances, the improvement on the performances are constantly on demand because of the rapid growth of data on web space. To satisfy the demand, many re-ranking methods, which enhance the performances by reordering retrieved results with independent algorithms, has been proposed. Conventional re-ranking algorithms are based on the assumption that visual patterns are not used on initial image retrieval stage. However, image search engines in present have begun to use the visual and the assumption is required to be reconsidered. Also, though it is possible to suspect that integration of multiple retrieval systems can improve the overall performance, the research on the topic has not been done sufficiently. In this paper, we made the condition that other manner than cooperation cannot improve the ranking result. We evaluate the algorithm on toy model and show that propose module can improve the retrieval results.

• Journal of Korea Multimedia Society
• /
• v.16 no.11
• /
• pp.1305-1314
• /
• 2013

The existing tag based system deducts a retrieval result with low accuracy through the usage of a single tag matching by using tags tagged in contents. And the system doesn't provide effectively contents related information which the tags have, as the users place tags on contents without considering the priority and associative relation between tags. For a solve of above problems, this paper suggests a tag ranking system which extracts semantic similarity between tags and re-ranks the tags tagged in contents. In order to evaluate the performance of suggested system, this paper experiments and compares the ranking result of this paper's tag ranking system with the result of baseline method using tags tagged in images and frequency method adapting tag co-appearance frequency.

• Journal of Korea Multimedia Society
• /
• v.14 no.11
• /
• pp.1491-1499
• /
• 2011

Now a days, to make good use of tags is a general tendency when users need to upload or search some multimedia data such as images and videos on the Web. In this paper, we introduce an approach to calculate semantic importance of tags and to make re-ranking with them on tagged Web image retrieval. Generally, most photo images stored on the Web have lots of tags added with user's subjective judgements not by the importance of them. So they become the cause of precision rate decrease with simple matching of tags to a given query. Therefore, if we can select semantically important tags and employ them on the image search, the retrieval result would be enhanced. In this paper, we propose a method to make image retrieval re-ranking with the key tags which share more semantic information with a query or other tags based on Wikipedia-based semantic relatedness. With the semantic relatedness calculated by using huge on-line encyclopedia, Wikipedia, we found the superiority of our method in precision and recall rate as experimental results.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.9
• /
• pp.59-68
• /
• 2022

A recommender system covers users, searches the items or services which users will like, and let users purchase them. Because recommendations from a recommender system are predictions of users' preferences for the items which they do not purchase yet, it is rarely possible to be drawn a perfect answer. An evaluation has been conducted to determine whether a prediction is right or not. However, it can be lower user's satisfaction if a recommender system focuses on only the preferences, that is caused by a 'filter bubble effect'. The filter bubble effect is an algorithmic bias that skews or limits the information an individual user sees on the recommended list. It is the reason why multiple metrics are required to evaluate recommender systems, and a diversity metrics is mainly used for it. In this paper, we compare three different methods for enhancing diversity for personalized recommendation - bin packing, weighted random choice, greedy re-ranking - with a practical e-commerce data acquired from a fashion shopping mall. Besides, we present the difference between experimental results and F1 scores.

• KIISE Transactions on Computing Practices
• /
• v.21 no.2
• /
• pp.144-147
• /
• 2015

Question Answering (QA) services can provide exact answers to user questions written in natural language form. This research focuses on how to build a QA system for a specific domain area. Online and offline QA system architecture of targeted domain such as domain detection, question analysis, reasoning, information retrieval, filtering, answer extraction, re-ranking, and answer generation, as well as data preparation are presented herein. Test results with an official Frequently Asked Question (FAQ) set showed 68% accuracy of the top 1 and 77% accuracy of the top 5. The contribution of each part such as question analysis system, document search engine, knowledge graph engine and re-ranking module for achieving the final answer are also presented.