BibSonomy publications for /user/hotho/mlWed Nov 30 14:45:00 CET 2011Proceedings of the ACM WebSci'11JuneTagging data as implicit feedback for learning-to-rank20112011 dm feedback learning logsonomy ml myown search social Thu Nov 17 14:21:19 CET 2011cite arxiv:1111.3689CBLOCK: An Automatic Blocking Mechanism for Large-Scale De-duplication
Tasks2011algorithm detection duplicate ml toread De-duplication---identification of distinct records referring to the same
real-world entity---is a well-known challenge in data integration. Since very
large datasets prohibit the comparison of every pair of records, {\em blocking}
has been identified as a technique of dividing the dataset for pairwise
comparisons, thereby trading off {\em recall} of identified duplicates for {\em
efficiency}. Traditional de-duplication tasks, while challenging, typically
involved a fixed schema such as Census data or medical records. However, with
the presence of large, diverse sets of structured data on the web and the need
to organize it effectively on content portals, de-duplication systems need to
scale in a new dimension to handle a large number of schemas, tasks and data
sets, while handling ever larger problem sizes. In addition, when working in a
map-reduce framework it is important that canopy formation be implemented as a
{\em hash function}, making the canopy design problem more challenging. We
present CBLOCK, a system that addresses these challenges. CBLOCK learns hash
functions automatically from attribute domains and a labeled dataset consisting
of duplicates. Subsequently, CBLOCK expresses blocking functions using a
hierarchical tree structure composed of atomic hash functions. The application
may guide the automated blocking process based on architectural constraints,
such as by specifying a maximum size of each block (based on memory
requirements), impose disjointness of blocks (in a grid environment), or
specify a particular objective function trading off recall for efficiency. As a
post-processing step to automatically generated blocks, CBLOCK {\em rolls-up}
smaller blocks to increase recall. We present experimental results on two
large-scale de-duplication datasets at Yahoo!---consisting of over 140K movies
and 40K restaurants respectively---and demonstrate the utility of CBLOCK.
Tue Jun 08 18:07:44 CEST 2010International Semantic Web Conferenceconf/semweb/2006640-653Lecture Notes in Computer ScienceTree-Structured Conditional Random Fields for Semantic Annotation.42732006annotation crf extraction information ml semantic Fri May 15 17:30:12 CEST 2009New York, NY, USAAIRWeb '08: Proceedings of the 4th international workshop on Adversarial information retrieval on the web61--68The Anti-Social Tagger - Detecting Spam in Social Bookmarking Systems20082008 bookmarking classification dm folksonomy mining ml myown social spam Wed May 06 11:51:59 CEST 2009Advances in Knowledge Discovery and Data Mining1-34From Data Mining to Knowledge Discovery: An Overview.1996definition dm kdd ml Sun May 03 14:48:41 CEST 2009New York, NY, USASIGMOD '93: Proceedings of the 1993 ACM SIGMOD international conference on Management of data207--216Mining association rules between sets of items in large databases1993association basic mining ml rules Mon Mar 23 21:50:27 CET 2009International Journal of Data Warehouse and Mining31--13Multi Label Classification: An Overview32007classification learning machine ml multi_label survey Nowadays, multi-label classification methods are increasingly required by modern applications, such as protein function classification, music categorization and semantic scene classification. This paper introduces the task of multi-label classification, organizes the sparse related literature into a structured presentation and performs comparative experimental results of certain multi-label classification methods. It also contributes the definition of concepts for the quantification of the multi-label nature of a data set.Fri Dec 19 09:19:20 CET 2008Amsterdam, The Netherlands, The NetherlandsApplied Ontology1-241--62AEON - An approach to the automatic evaluation of ontologies320082008 automatic evaluation ml myown ontology sw OntoClean is an approach towards the formal evaluation of taxonomic relations in ontologies. The application of OntoClean consists of two main steps. First, concepts are tagged according to meta-properties known as rigidity, unity, dependency and identity. Second, the tagged concepts are checked according to predefined constraints to discover taxonomic errors. Although OntoClean is well documented in numerous publications, it is still used rather infrequently due to the high costs of application. Especially, the manual tagging of concepts with the correct meta-properties requires substantial efforts of highly experienced ontology engineers. In order to facilitate the use of OntoClean and to enable the evaluation of real-world ontologies, we provide AEON, a tool which automatically tags concepts with appropriate OntoClean meta-properties and performs the constraint checking. We use the Web as an embodiment of world knowledge, where we search for patterns that indicate how to properly tag concepts. We thoroughly evaluated our approach against a manually created gold standard. The evaluation shows the competitiveness of our approach while at the same time significantly lowering the costs. All of our results, i.e. the tool AEON as well as the experiment data, are publicly available.Mon Dec 01 15:38:40 CET 2008New York, NY, USACIKM '08: Proceeding of the 17th ACM conference on Information and knowledge mining93--102A sparse gaussian processes classification framework for fast tag suggestions2008bibsonomy bookmarking classification dataset ml recommender social tag tagging taggingsurvey toread Thu Sep 11 12:03:43 CEST 2008String Processing and Information Retrieval98--109The Intention Behind Web Queries2006analysis dm intention ml query search toread The identification of the user’s intention or interest through queries that they submit to a search engine can be very useful
to offer them more adequate results. In this work we present a framework for the identification of user’s interest in an automaticway, based on the analysis of query logs. This identification is made from two perspectives, the objectives or goals of auser and the categories in which these aims are situated. A manual classification of the queries was made in order to havea reference point and then we applied supervised and unsupervised learning techniques. The results obtained show that fora considerable amount of cases supervised learning is a good option, however through unsupervised learning we found relationshipsbetween users and behaviors that are not easy to detect just taking the query words. Also, through unsupervised learning weestablished that there are categories that we are not able to determine in contrast with other classes that were not consideredbut naturally appear after the clustering process. This allowed us to establish that the combination of supervised and unsupervisedlearning is a good alternative to find user’s goals. From supervised learning we can identify the user interest given certainestablished goals and categories; on the other hand, with unsupervised learning we can validate the goals and categories used,refine them and select the most appropriate to the user’s needs.Wed Feb 13 12:26:23 CET 2008Künstliche Intelligenz35-8Mining the World Wide Web20072007 introduction ir ki mining ml myown web Mon Jan 14 14:47:18 CET 2008Informatica2141-141Introduction to the Special Issue 'Learning in Web Search'3020062006 ir ml search web Mon Jan 14 09:58:03 CET 2008AUGWorkshop at the 22nd International Conference on Machine Learning (ICML 2005) Proceedings of the Workshop on Learning in Web Search (LWS 2005) 20052005 ir ml myown search web Fri Sep 28 16:05:05 CEST 2007Probabilistic Robotics (Intelligent Robotics and Autonomous Agents)2001dm learning machine ml probabilistic Tue Jun 05 09:54:45 CEST 2007JULFrontiers in Artificial IntelligenceOntology Learning from Text: Methods, Evaluation and Applications1232005dm learning ml ol semantic web Fri May 18 19:27:34 CEST 2007Modeling the Internet and the Web: Probabilistic Methods and AlgorithmsAprilModeling the Internet and the Web: Probabilistic Methods and Algorithms2003KI2007WebMining dm kdd mining ml web Modeling the Internet and the Web covers the most important aspects of modeling the Web using a modern mathematical and probabilistic treatment. It focuses on the information and application layers, as well as some of the emerging properties of the Internet. Provides a comprehensive introduction to the modeling of the Internet and the Web at the information level. Takes a modern approach based on mathematical, probabilistic, and graphical modeling. Provides an integrated presentation of theory, examples, exercises and applications. Covers key topics such as text analysis, link analysis, crawling techniques, human behaviour, and commerce on the Web. Interdisciplinary in nature, Modeling the Internet and the Web will be of interest to students and researchers from a variety of disciplines including computer science, machine learning, engineering, statistics, economics, business, and the social sciences.Wed Jan 17 15:56:27 CET 2007SIGIR Forum{T}he {W}ikipedia {X}{M}{L} {C}orpus2006data dm mining xml corpus ml wikipedia Fri Dec 15 08:08:45 CET 2006{C4.5 Programs for Machine Learning}1993learning decision entropy ***** tree ml Fri Oct 27 11:56:43 CEST 2006ACM Computing Surveys11--47Machine learning in automated text categorization342002tm text survey classification categorization ml Fri Jun 23 07:06:19 CEST 2006{UCI} Repository of machine learning databases1998learning data dataset dm mining machine ml uci