PirateBay Download | kopotek m language processing intelligent information systems 2013

kopotek m language processing intelligent information systems 2013

Type:: E-books
Files:: 1
Size:: 6.9 MB
Uploaded On:: March 30, 2024, 1:07 p.m.
Added By:: andryold1
Seeders:: 6
Leechers:: 2
Info Hash:: 0FF6520A424367A77CC085444D3CC8F2B5B38D89

Textbook in PDF format The Conference on Intelligent Information Systems, organized by the Institute of Computer Science of the Polish Academy of Sciences, has been a prominent meeting place for scientists from all over the world for nearly 30 years now. This volume contains the papers presented at the 20th conference of the series Language Processing and Intelligent Information Systems Conference, which was held in Warsaw, Poland, June 17 – 18, 2013. This year, the main goal of the meeting was to present new ideas and tools for natural language processing and interplay between the analysis of natural language and traditional machine learning technologies. The volume contains 28 papers selected by the Program Committee from 53 submitted papers. Each paper was anonymously reviewed by at least two independent reviewers, and the articles presented in this volume were significantly improved on the basis of the reviewers’ comments. The volume consists of two types of submissions: 13 long research papers describing original research contributions to the conference topics, and 15 short papers mainly describing tools and resources. Two demonstration sessions were organized to take place during the conference. One of these sessions presented the results of the four-year Polish POIG.01.01.02 project “An adaptive system to support problem-solving on the basis of document collections on the Internet.” A short description of the tools and systems to be presented are published on the conference website. The papers presented in this volume are organized into three thematic groups: natural language processing, clustering and classification of big collections of textual documents, and classic data mining problems. The conference was opened by an invited plenary talk entitled “Syntax and Semantics in the Web-Scale Extraction of n-ary Relations” by Hans Uszkoreit — Scientific Director at the German Research Center for Artificial Intelligence (DFKI) and Head of the DFKI Language Technology Lab, Saarbr¨ucken, Germany. The second day started with a talk “Beyond Query Suggestions: Recommending Tasks to SE Users” given by Fabrizio Silvestri from the Institute of Information Science and Technologies at Consiglio Nazionale delle Ricerche, Pisa, Italy. In the talk, he discussed the Task Relation Graph (TRG) as a representation of users’ search behaviors on a task-by-task perspective. The problems of natural language processing are discussed in 17 articles. These papers concern the processing of a set of very diverse languages: Polish, Croatian, English, German, Dutch, Chinese, and Persian. Both machine learning and formal approaches are explored and hybrid solutions combining more than one technique are frequently proposed. A hybrid approach is used for multilingual toponyms extraction, for example. In the paper by M. Habib and M. van Keulen, an HMM module serves for the selection of potential candidates for toponyms, while the disambiguation level is done using CRF. For many languages, the tools for performing basic syntactic analysis are not well developed. Relatively free word order and rich inflection render some of the methods used for defining English or German grammar unsatisfactory in dealing with these languages. A great number of morphological tags and the relatively small size of annotated corpora make this task more challenging. Among papers dealing with this issue is the work of K. Krasnowska, which describes a Polish LTAG grammar, while two other papers represent a partial parsing approach. A. Radziszewski and A. Pawlaczek describe recognition of CRF-based chunks, while A. Radziszewski et al. use C4.5, SVM and a memory-based classifier to classify predicate-argument relationships. A. Wroblewska and P. Sikora present an on-line service of the newly established Polish dependency parser. Papers presenting specific NLP applications concern recognition of fake reviews (M. Rubikowski and A. Wawer), question answering (P. Przybyła), and recognition of named entities (A. L.-F. Han et al). Another group of papers is devoted to resource building, evaluation, and sharing. These works comprise, among others, the creation of a Croatian derivational dictionary in a paper by V. ˇStefanec et al., the annotating of named entities by E. Hajnicz, and the detection of annotation errors in existing treebanks or corpora addressed in two papers: one by Ł. Kobyliński and another by K. Krasnowska and A. Przepiorkowski. M. Aminian et al. describe a spectral clustering algorithm used for identification of Persian semantic verb classes based on syntactic information, M. Ogrodniczuk describes a translational – based co-reference resolution for Polish, while M. Marcińczuk and A. Radziszewski present a general language for text annotation. Two papers deal with processing older language variants. X. Zou et al. describe a method for recognizing changes in usage of a word in time on the basis of the context of its occurrences, while J. Waszczuk presents an architecture of a dictionary of old Polish. The authors of seven articles included in the second part of the volume extend traditional tasks of machine learning, such as clustering or classification, to the domain of collections of textual documents G. Stratogiannis et al. investigate the issue of reliable search for related entities using semantic knowledge extracted from Wikipedia. With such knowledge a semantic relatedness between entities is established, and, finally, a semantic clustering is used to answer a given question. R. Szmit presents an algorithm and technological framework for the search for similar documents based on locally sensitive hashing. The proposed distributed algorithm is designed to cope with very large document collections. M. Dramiński et al. look at the clustering of user activity data from various topically related sites as a vehicle for obtaining better user profiles. Although the idea seems to be plausible, the authors demonstrate that we are far from being able to apply it in practice as the users fluctuate between the clusters. T. Kuśmierczyk and M. Sydow reiterate the old problem of focused Web crawling. They demonstrate that the usage of short lists of keywords, shallow search, and appropriately chosen starting pages may dramatically improve the Harvest Ratio. M. Łukasik and M. Sydow investigate the properties of a version of the multi-label classification algorithm based on the k-Nearest Neighbors method. They show that the modification, concentrating on choosing appropriate thresholding, performs significantly better than the standard form of the algorithm. T. Giannakopoulos et al. apply a supervised learning technique for classifying documents in a manner that allows visualization of the contents of a collection of scientific documents. M. A. Kłopotek et al. turn to the issue of balance between personalization and the required space for storing data. They propose a method for combining a personalized PageRank, computed for various categories, to obtain a PageRank for a joint category, so that a considerable number of ranking vectors need not be stored. The last group of contributors reports on new results obtained in the domain of classic data mining. M. Lucińska and S.T. Wierzchoń propose a new spectral clustering algorithm that uses a novel way of identifying the cluster number solely on the basis of the eigenvector structure. They demonstrate that the approach yields valid clusters, even in the case of data sets that are not clearly cut. R. Kłopotek investigates a recently proposed generator of artificial social graphs with a bipartite structure. He seeks to reconstruct generator parameters from the generated graphs while posing the question of why such generators are able to provide artificial graphs that behave similarly to real ones. K. Trojanowski and M. Janiszewski investigate the issue of the influence of resource constraints on the outcome of an optimization algorithm. They introduce the concept of user impatience and demonstrate its impact on the expected value of the result of optimization. C. Sur et al. propose an algorithm for solving the traveling salesman problem that exploits new nature-based techniques of local search. We would like to express our thanks to the invited speakers and the authors of the papers for their contribution. Likewise we thank the authors of the demonstrated systems. We extend special thanks to all the members of the Program Committee and invited reviewers for their excellent job

Get This Torrent

Kłopotek M. Language Processing and Intelligent Information Systems 2013.pdf 6.9 MB

Details for: Kłopotek M. Language Processing and Intelligent Information Systems 2013

kopotek m language processing intelligent information systems 2013