izpis_h1_title_alt

Oblikoskladenjsko označevanje slovenskega jezika z globokimi nevronskimi mrežami
Belej, Primož (Author), Robnik Šikonja, Marko (Mentor) More about this mentor... This link opens in a new window, Krek, Simon (Co-mentor)

.pdfPDF - Presentation file, Download (4,95 MB)

Abstract
V magistrskem delu se ukvarjamo z oblikoskladenjskim označevanjem slovenskega jezika. Pri tej nalogi s področja obdelave naravnega jezika povedim priredimo ustrezno zaporedje oznak, ki opisujejo oblikoskladenjske lastnosti besed. Za razliko od tipičnih pristopov, ki vhodne povedi obravnavajo na nivoju besed, naša rešitev obravnava vhodne povedi kot zaporedja znakov. Nalogo označevanja rešujemo s kombinacijo konvolucijskih in rekurentnih nevronskih mrež. Posebnost našega pristopa je tudi v sami naravi označevanja, saj ga ne obravnavamo kot problem večrazredne klasifikacije, temveč kot večznačno klasifikacijo, kjer primerom dodeljujemo oznake. Z namenom izboljšave rezultatov našo rešitev združimo v ansambel treh označevalnikov, skupaj z dvema obstoječima označevalnikoma za slovenski jezik. Ob primerjavi naše rešitve z obstoječimi ugotovimo, da predlagana rešitev dosega najboljše rezultate pri reševanju zadanega problema.

Language:Slovenian
Keywords:strojno učenje, oblikoskladenjsko označevanje, globoko učenje, konvolucijske nevronske mreže, rekurentne nevronske mreže, ansambli klasifikatorjev
Work type:Master's thesis/paper (mb22)
Organization:FRI - Faculty of computer and information science
Year:2018
Views:127
Downloads:85
Metadata:XML RDF-CHPDL DC-XML DC-RDF
 
Average score:(0 votes)
Your score:Voting is allowed only to logged in users.
:
Share: Bookmark and Share

Secondary language

Language:English
Title:Part of speech tagging of slovene language using deep neural networks
Abstract:
The thesis deals with part of speech tagging of Slovene language. Part of speech tagging is a process of matching sentences in natural language with a sequence of suitable tags, which contain information about parts of speech and morphological properties of words. Our solution uses character-level representation of words, which is different from typical solutions, which process input sentences as sequences of words. Our part of speech tagger is implemented using convolutional and recurrent neural networks. Unlike common approaches that address this problem as multi-class classification, our solution proposes a multi-label classification approach. In order to improve our results we implement an ensemble of three part of speech taggers. When comparing our solution with existing ones, we find that the proposed solution achieves the best results.

Keywords:machine learning, part-of-speech tagging, deep learning, convolutional neural networks, recurrent neural networks, ensemble classifiers

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back