Nadzirano učenje u strojnom provjerniku pravopisa

undergraduate thesis

• Title (english): Supervised learning in an online spellchecking service
• Author: Nikša Brala
• Mentor: Gordan Gledec (mentor)
• Committee member: Gordan Gledec (predsjednik povjerenstva)
• Committee member: Bartol Boras (član povjerenstva)
• Committee member: Kristijan Poje (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-02, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Nadzirano učenje u strojnom provjerniku pravopisa
Strojni provjernik pravopisa Hascheck, danas poznatiji po svom web-sjedištu Ispravi.me, svake godine broji sve više posjeta i obrađenih tekstova. Provjernik pritom neprestano svim novim i nepoznatim, ali smislenim, riječima koje obradi nadziranim učenjem širi i razvija svoj rječnik. Nadzor je potreban radi očuvanja točnosti rječnika, pa ulazak svake riječi u rječnik mora odobriti čovjek, a riječi koje mu dolaze na ručnu provjeru poželjno je što bolje pročistiti od nesuvislih i pogrešnih riječi tako da se ljudski rad svede na minimum. Ovaj rad nudi programsko rješenje za izdvajanje glagolskih oblika iz korpusa prijavljenih grešaka. Programsko se rješenje temelji na jednostavnoj i programski ostvarivoj klasifikaciji glagola hrvatskog jezika po nastavcima koja je iscrpno opisana i obrađena kroz poglavlja ovog rada. Uz klasični programerski pristup, detaljno je opisan i pristup izrade programskog rješenja temeljenog na metodi nadziranog strojnog učenja - stablu odluke. Na kraju su rada prikazani rezultati i uspješnost primjene ostvarenog programskog rješenja zajedno s opisom varijabilnog ulaznog parametara koji je moguće jednostavno prilagoditi ulaznom skupu podataka.

Abstract (english)

Supervised learning in an online spellchecking service
An online spellchecking service Hascheck, today better known for its new website Ispravi.me, continues its trend of rapid growth thanks to the increasing number of total website visits and total number of processed tasks. In doing so, the spellchecker tends to continuously expand and develop its dictionary with all new and unknown but meaningful words it processes through supervised learning. Supervision is necessary to maintain the accuracy of the dictionary, so the entry of each word into the dictionary must be approved by a human. It is desirable to filter out nonsensical and incorrect words as much as possible before they reach manual verification, thus minimizing human effort. This paper offers a programmatic solution for extracting verb forms from the corpus of reported errors. The solution is based on a simple and programmatically feasible classification of verbs in Croatian language by their suffixes, which is comprehensively described and elaborated upon in the chapters of this paper. Alongside the classical programming approach, the paper also provides details for the development of the programmatic solution based on the supervised machine learning method - decision tree. At the end of the paper, the results and effectiveness of the implemented programmatic solution are presented, along with a description of the variable input parameter that can be easily adjusted to the input data set.

• Keywords: Hascheck; Ispravi.me; Strojni provjernik pravopisa; Hrvatski jezik; Glagoli; Stablo odluke
• Keywords (english): Hascheck; Ispravi.me; Spellchecker; Croatian Language; Verbs; Decision tree
• Language: croatian
• URN:NBN: urn:nbn:hr:168:198077
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 20:10:48