|
|
Přibližná extrakce frázové tabulky z velkého paralelního korpusu
Przywara, Česlav ; Bojar, Ondřej (advisor) ; Tamchyna, Aleš (referee)
The aim of this work is to examine the applicability of an algorithm for approximate frequency counting to act as an on-the-fly filter in the process of phrase table extraction in Statistical Machine Translation systems. Its implementation allows for the bulk of extracted phrase pairs to be much reduced with no significant loss to the ultimate quality of the phrase-based translation model as measured by the state-of-the-art evaluation measure BLEU. The result of this implementation is a fully working program, called eppex, capable of acting as an alternative to the existing tools for phrase table creation and filtration that are part of the open-source SMT system Moses. A substantial part of this work is devoted to the benchmarking of both the runtime performance and the quality of produced phrase tables achieved by the program when confronted with parallel training data comprised of 2 billions of words. Powered by TCPDF (www.tcpdf.org)
|
|
|
Methods of multiword expression extraction from text
Przywara, Česlav ; Pecina, Pavel (advisor) ; Schlesinger, Pavel (referee)
The goal of this thesis is an effective implementation of the methods of multiword expression extraction from text, so that designed program would be capable of processing large textual corpora containing up to billions of words. Additional function of the program is context tracing of extracted N-grams. For thesis purposes the program implementation is specially adjusted for collocation extraction from The Prague Dependency Treebank, but the program is designed in such manner that allows an easy future extensibility.
|
|
|
Přibližná extrakce frázové tabulky z velkého paralelního korpusu
Przywara, Česlav ; Bojar, Ondřej (advisor) ; Tamchyna, Aleš (referee)
The aim of this work is to examine the applicability of an algorithm for approximate frequency counting to act as an on-the-fly filter in the process of phrase table extraction in Statistical Machine Translation systems. Its implementation allows for the bulk of extracted phrase pairs to be much reduced with no significant loss to the ultimate quality of the phrase-based translation model as measured by the state-of-the-art evaluation measure BLEU. The result of this implementation is a fully working program, called eppex, capable of acting as an alternative to the existing tools for phrase table creation and filtration that are part of the open-source SMT system Moses. A substantial part of this work is devoted to the benchmarking of both the runtime performance and the quality of produced phrase tables achieved by the program when confronted with parallel training data comprised of 2 billions of words. Powered by TCPDF (www.tcpdf.org)
|
|
|
Methods of multiword expression extraction from text
Przywara, Česlav ; Schlesinger, Pavel (referee) ; Pecina, Pavel (advisor)
The goal of this thesis is an effective implementation of the methods of multiword expression extraction from text, so that designed program would be capable of processing large textual corpora containing up to billions of words. Additional function of the program is context tracing of extracted N-grams. For thesis purposes the program implementation is specially adjusted for collocation extraction from The Prague Dependency Treebank, but the program is designed in such manner that allows an easy future extensibility.
|
guest ::
