{"id":157,"date":"2006-11-28T13:13:00","date_gmt":"2006-11-28T04:13:00","guid":{"rendered":"https:\/\/fugutabetai.com\/blog\/2006\/11\/28\/international-conference-on-asian-digital-libraries-icadl2006-in-kyoto-day-2\/"},"modified":"2006-11-28T13:13:00","modified_gmt":"2006-11-28T04:13:00","slug":"international-conference-on-asian-digital-libraries-icadl2006-in-kyoto-day-2","status":"publish","type":"post","link":"https:\/\/fugutabetai.com\/blog\/2006\/11\/28\/international-conference-on-asian-digital-libraries-icadl2006-in-kyoto-day-2\/","title":{"rendered":"International Conference on Asian Digital Libraries (ICADL2006) in Kyoto day 2"},"content":{"rendered":"<ul>\n<li><a href=\"#day2harvesting\">A Digital Resource Harvesting Approach<br \/>\nfor Distributed Heterogeneous Repositories<\/a><\/li>\n<li><a href=\"#day2parallel\">Parallelising Harvesting<\/a><\/li>\n<li><a href=\"#day2personalized\">Personlized Information Delivering<br \/>\nService in Blog-like Digital Libraries<\/a><\/li>\n<li><a href=\"#day2ontology\">A Personal Ontology Model for Library<br \/>\nRecommendation System<\/a><\/li>\n<li><a href=\"#day2recommendation\">Research and Implementation of a<br \/>\nPersonalized Recommendation System<\/a><\/li>\n<li><a href=\"#day2imagespotting\">Text Image Spotting Using Local<br \/>\nCrowdedness and Hausdorff Distance<\/a><\/li>\n<li><a href=\"#day2effective\">Effective Image Retrieval for the<br \/>\nM-learning System<\/a><\/li>\n<li><a href=\"#day2mediatranslation\">Language Translation and Media<br \/>\nTransformation in Cross-Language Image Retrieval<\/a><\/li>\n<li><a href=\"#day2surfaceerrors\">A Surface Errors Locator System for<br \/>\nAncient Chinese Culture Preservation<\/a><\/li>\n<\/ul>\n<p><!-- readmore --><\/p>\n<p><a name=\"day2harvesting\"><\/p>\n<h4>&#8220;A Digital Resource Harvesting Approach for Distributed<br \/>\nHeterogeneous Repositories&#8221;, <i>Yang Zhao, Airon Jiang<\/i> (Tsinghua<br \/>\nUniversity Library, China)<\/h4>\n<p><\/a><\/p>\n<p>They present a digital resource harvesting system that uses a few<br \/>\ndifferent standards (which I don&#8217;t really know about) to get around<br \/>\nproblems in using plain HTTP.  Their messaging system is able to more<br \/>\neasily transfer large data payloads.  They use METS (metadata encoding<br \/>\nand transmission standard) which is supported by some digital library<br \/>\nfederation.  They have tested their system in CALIS-ETD, a large<br \/>\nChinese consortium for distributing electron thesis and dissertation.  <\/p>\n<p><a name=\"day2parallel\"><\/p>\n<h4>\n&#8220;Parallelising Harvesting&#8221;, <i>Hussein Suleman<\/i> (University of Cape<br \/>\nTown, South Africa)<br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>For two open source systems, EPrint and DSpace, they both become quite<br \/>\nslow with over 100,000 documents.  In pure sciences (Physics,<br \/>\nChemistry, etc.) they have lots of data and use high performance<br \/>\ncomputer systems to handle the large sets of data.  He explains some<br \/>\ntypes of large computer systems (clusters, multi-core, grid, etc.)  He<br \/>\nused OpenMosix open source system to create a cluster.  <\/p>\n<p><P\/> <\/p>\n<p>For harvesting there is the standard OAI Protocol for Metadata<br \/>\nharvesting transfers.  Since that is a serial protocol, there is a<br \/>\ntoken passed around the grid and only the token owning node can<br \/>\nrequest data.  The rest process data.  He did some experiments on<br \/>\ndifferent architectures impacts on data transfer and data indexing.<br \/>\nIn the end, if you have more computation to do on the data than data<br \/>\ntransfer, the cluster is a win.  <\/p>\n<p><P\/> Also, the OAI protocol needs to be extended to allow for parallel<br \/>\naccess.  <\/p>\n<h3>Session 5a: Personalization in Digital Libraries<\/h3>\n<p><a name=\"day2personalized\"><\/p>\n<h4>\n&#8220;Personlized Information Delivering Service in Blog-like Digital<br \/>\nLibraries&#8221;, <i>Jason J. Jung (Inha University, Korea; INRIA<br \/>\nRhone-Alpes, France.)<\/i><br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>In the Blogosphere there is a problem of <i>information overload<\/i>,<br \/>\nit is difficult to find relevant information.  Another problem is the<br \/>\nnetwork isolation phenomenon.  The network space is generally a<br \/>\npersonal set of links.  The speaker&#8217;s BlogGrid is a method to approach<br \/>\nthese problems.  They define and collect user activities, extract user<br \/>\npreferences, and then make recommendations to the users.  The system<br \/>\nfollows users&#8217; posts, their linking patterns, their navigation<br \/>\npatterns (Random or access to neighbors), their responses (trackbacks,<br \/>\ncomments), and blog categories (using ODP or TopicMap taxonomies.)  <\/p>\n<p><P\/> <\/p>\n<p>When users exhibit similar response patterns, we can assume that they<br \/>\nexhibit the same interests.  The speaker wants to find &#8220;virtual hub&#8221;,<br \/>\nso they use the distance between users, and hub weight which looks to<br \/>\nbe similar to the Google algorithm (since the equation has authorities<br \/>\nin there too.)  <\/p>\n<p><P\/><\/p>\n<p>They had a user evaluation where they asked students to track new<br \/>\ninformation, some using BlogGrid, some not.  Most liked the<br \/>\nrecommendations from the users, but some people did not like the<br \/>\nsystem due to having to install it on their systems and privacy<br \/>\nissues.  <\/p>\n<p><a name=\"day2ontology\"><\/p>\n<h4>\n&#8220;A Personal Ontology Model for Library Recommendation System&#8221;, <i>I-En<br \/>\nLiao (National Chung Hsing University, Taiwan), Shu-Chuan Liao (Asia<br \/>\nUniversity, Taiwan), Kuo-Fong Kao,<br \/>\nIne-Fei Harn (National Chung Hsing University)<\/i><br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>There are two approaches: social filtering (other people who liked X<br \/>\nalso liked Y), and content-based filtering.  This might not be as<br \/>\napplicable to libraries because library users are not as interested in<br \/>\npopularity; they have a specific information need.  Their paper looks<br \/>\nat content-based recommendation.  <\/p>\n<p><P\/><\/p>\n<p>Their objectives are to automatically mine user interests from loan<br \/>\nrecord, re-rank keywords based on a user interest score from a<br \/>\npersonal ontology.  Their system is a web-based application.  When the<br \/>\nuser logs in, their record is analyzed, then books are recommended to<br \/>\nthem.  Their reference ontology could be Library of Congress (LCC) or<br \/>\na Chinese ontology (CCL).  They build a personal ontology by using the<br \/>\nloan record to find favorite categories from the reference ontologies.<br \/>\nThey copy over the nodes that have a value greater than some<br \/>\nthreshold.  Then they try to find interesting keywords for the user.<br \/>\nThey compute a kind of TF*IDF over the keywords and then fetch books<br \/>\nbased on the value of their highest keywords (I&#8217;m surprised they<br \/>\ndidn&#8217;t do some sort of weighting function, they just took the value of<br \/>\nthe highest rated keyword!)<\/p>\n<p><a name=\"day2recommendation\"><\/p>\n<h4>\n&#8220;Research and Implementation of a Personalized Recommendation System&#8221;,<br \/>\n<i>Li Dong (Tsinghua University, China), Yu Nie (SINA Corporation,<br \/>\nChina), Chunxiao Xing, Kehong Wang (Tsinghau University China)<\/i>.<br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>They have a clustering algorithm (static that runs once, then dynamic<br \/>\nfor interacting with the user.)  Their &#8220;cluster mining&#8221; approach<br \/>\nallows data elements to overlap.  (There are standard approaches in<br \/>\nclustering that allow for that as well.)  They did a clustering<br \/>\nevaluation by looking at how many of the items in the cluster were<br \/>\nrecommended, but I don&#8217;t know if they have a user evaluation included<br \/>\nhere.  It looks like they also did some user evaluation that looks at<br \/>\nhow often the users took the recommendations, but I&#8217;m really not too<br \/>\nsure about that.  They used some data from a website where they are<br \/>\nrunning their recommendation system.  <\/p>\n<p><P\/><\/p>\n<p>Overall for the talks this morning, I feel like I should read the<br \/>\npapers because in general the talks were not very good.  Sometimes the<br \/>\npresentation language (English) was a bit troublesome, other times I<br \/>\njust didn&#8217;t get a feel that the talk was well organized.  Hussein<br \/>\nSuleman&#8217;s talk was nice, but he was aiming for more the librarian end<br \/>\nof things, and was a bit simplified.  <\/p>\n<hr>\n<h3>Session 7c: Multimedia Resource Retrieval and Organization<\/h3>\n<p><a name=\"day2imagespotting\"><\/p>\n<h4>\n&#8220;Text Image Spotting Using Local Crowdedness and Hausdorff Distance&#8221;,<br \/>\n<i>Hwa-Jeong Son, Sang-Cheol Park, Soo-Hyung Kim, Ji-Soo Kim, Guue Sang<br \/>\nLee, Deok Jai Choi (Chonnam National University, Korea)<\/i><br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>They are looking at spotting text in an image given a query.  They are<br \/>\ntaking the query as an image, instead of as text.  They are trying to<br \/>\nmatch the query image to a sub-region of the document image.  They use<br \/>\nthe Hausdorff distance to match the images, and tried two other<br \/>\napproaches: binary correlation and a modified Hausdorff distance.  <\/p>\n<p><P\/> Binary correlation just looks at an average of pixels in an<br \/>\noverlapping region of the those images.  You can move the image around<br \/>\nto find the minimum distance over the entire map.  (That seems quite<br \/>\ncomputationally expensive!!)  They also modify Hausdorff&#8217;s approach to<br \/>\nmake it less computationally expensive by looking at &#8220;local<br \/>\ncrowdedness&#8221; &#8211; which looks to me like doing a blur operationand the<br \/>\ntaking points that are heavily dark from the bleed-over from adjacent<br \/>\npixels.  <\/p>\n<p><P\/>  They build probability distributions for class given the<br \/>\nfeatures, and then classify based on the probability distributions<br \/>\ngiven the features.  <\/p>\n<p><P\/>  They test over 380 documents with 100%, 70%, 50%?, and 30%? of<br \/>\nthe query in the document &#8212; but also half of those were used for<br \/>\ntraining.  So that is a very large document to query<br \/>\nsize I think.  Their proposed method performed best.  <\/p>\n<p><P\/>  For future research they are looking at scaling and font<br \/>\nvariations, and would like to look at newer faster features.   <\/p>\n<p><a name=\"day2effective\"><\/p>\n<h4>\n&#8220;Effective Image Retrieval for the M-learning System&#8221;,<br \/>\n<i>EunJung Han, AnJin Park, DongWuk Kyoung (Soongsil University,<br \/>\nKorea), HwangKyu Yang (Dongseo University, Korea), KeeChul Jung<br \/>\n(Soongsil University Korea).<\/i><br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>They are looking at blended learning environments where physical<br \/>\nlearning environments are augmented with virtual information on a PDA<br \/>\n(camera to capture marks, screen to display information.)  So they<br \/>\nwant to recognize real images instead of &#8220;pattern markers&#8221; and also to<br \/>\nadapt their algorithms to the low-computational resources of PDAs.  <\/p>\n<p><P\/>  They assume that the background is white, and the object to<br \/>\nrecognize is located in a central region.  They propose some rotation<br \/>\nand scale invariant features.  They do boundary extraction, and then<br \/>\ntake the starting point as the closet pixel to the centroid of the<br \/>\nobject.  That takes care of rotation.  There is a problem when the<br \/>\nclosest boundary pixel is on a circle though, so they use the closest<br \/>\n3 pixels.  They also use Dynamic Time Warping to compare images, which<br \/>\nwas much faster (a factor of 10) than an HMM approach.  <\/p>\n<p><P\/>  Their approach can have problems when the shapes have similar<br \/>\nboundaries, or when the boundaries are not extracted well.  They had<br \/>\n87% recognition rate over 30 objects in their DB.  <\/p>\n<p><a name=\"day2mediatranslation\"><\/p>\n<h4>\n&#8220;Language Translation and Media Transformation in Cross-Language Image<br \/>\nRetrieval&#8221;,<br \/>\n<i>Hsin-Hsi Chen, Yih-Chen Chang (National Taiwan University, Taiwan)<br \/>\n<\/i>.<br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>For cross-media information retrieval, queries and documents are in<br \/>\ndifferent media.  In this paper they are dealing with cross-language<br \/>\nimage retrieval.  In their format the images can be annotated in<br \/>\nmultiple languages.  They build a transmedia dictionary by looking at<br \/>\ntagged images, breaking it up into bits, and assigning visual features<br \/>\nto the image tags.  <\/p>\n<p><P\/>  They make a kind of comparable corpus by doing first content<br \/>\nbased retrieval, and then take the retrieved images and use their tags<br \/>\nto build a text-based query.  They present several methods for doing<br \/>\nthe translation at different times with different information<br \/>\n(relevance feedback or not, include query translation or not, merging<br \/>\nresults in different ways, etc.)  Adding the content information<br \/>\nimproves performance over just doing content-based IR.  <\/p>\n<p><a name=\"day2surfaceerrors\"><\/p>\n<h4>\n&#8220;A Surface Errors Locator System for Ancient Chinese Culture<br \/>\nPreservation&#8221;,<br \/>\n<i><br \/>\nYimin Yu, Duanqing Xu, Chun Chen, Yijun Yu, Lei Zhao (Zhejiang<br \/>\nUniversity, China).<br \/>\n<\/i><br \/>\n<\/h4>\n<p><\/a><\/p>\n<p>I think this talk was cancelled because the speaker was not here.<\/p>\n<hr>\n<p>The evening banquet is at<br \/>\n<a href=\"http:\/\/thegardenorientalkyoto.com\/eng\/\">The Garden Oriental<br \/>\nKyoto<\/a>.  <\/p>\n","protected":false},"excerpt":{"rendered":"<p>A Digital Resource Harvesting Approach for Distributed Heterogeneous Repositories Parallelising Harvesting Personlized Information Delivering Service in Blog-like Digital Libraries A Personal Ontology Model for Library Recommendation System Research and Implementation of a Personalized Recommendation System Text Image Spotting Using Local Crowdedness and Hausdorff Distance Effective Image Retrieval for the M-learning System Language Translation and Media [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[10],"tags":[],"_links":{"self":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts\/157"}],"collection":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/comments?post=157"}],"version-history":[{"count":0,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts\/157\/revisions"}],"wp:attachment":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/media?parent=157"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/categories?post=157"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/tags?post=157"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}