Number of the records: 1
Neural Network Based Boolean Factor Analysis: Efficient Tool for Automated Topics Search.
- 1.0032258 - ÚI 2007 RIV JO eng C - Conference Paper (international conference)
Húsek, Dušan - Frolov, A. A. - Polyakov, P.Y. - Řezanková, H.
Neural Network Based Boolean Factor Analysis: Efficient Tool for Automated Topics Search.
[Neurosíťová booleovská faktorová analýza: efektivní nástroj pro automatické vyhledávání témet.]
Computer Science and Information Technology. Amman: Applied Science Private University, 2006 - (Issa, G.; Qaddoura, E.; El-Qawasmeh, E.; Raho, G.), s. 321-327. ISBN 9957-8592-0-X.
[CSIT 2006. International Multiconference on Computer Science and Information Technology /4./. Amman (JO), 05.04.2006-07.04.2006]
R&D Projects: GA AV ČR 1ET100300419
Institutional research plan: CEZ:AV0Z10300504
Keywords : Boolean factor analysis * neural networks * associative memory * clustering * web searching * semantic web * information retrieval * document indexing * document classification * document processing * data mining * machine learning
Subject RIV: BB - Applied Statistics, Operational Research
The paper describes an automatic document concepts searching metod based on recurrent neural network implementation of Boolean factor analysis procedure. Advantage of this approach is the ability of effective analysis of large natural language databases, with rich vocabulary and easy concepts update. Hoppfield-like associative memory with parallel dynamics was substantionaly modified to fulfill this task. We developed totally new recall procedure that allows for the search of all attractors corresponding to factors (a true attractor). Necessary separation of spurious attractors is based on calculation of their Lyapunov function. Being applied to textual data the procedure allows to reveal groups of highly correlated words (factors) which frequently occur in documents jointly and represent concepts covered by these documents.
Uvedena je nová metoda pro automatické vyhledávání konceptů v textových databázích založená na rekurentní neuronové síti Hoppfieldova typu implementující Booleovou faktorovou analýzu. Výhodou tohoto přístupu je schopnost efektivní analýzi v rozsahlých databázích v přirozeném jazyce , s rozsáhlým slovníkem termů a konceptem snadné aktualizace. Nová asociativní paměť Hoppfieldova typu s paralelní dynamikou byla implementována pro řešení této úlohy
Permanent Link: http://hdl.handle.net/11104/0132818
Number of the records: 1