Automatsko prepoznavanje i normalizacija vremenskih izraza u nestrukturiranim novinskim i medicinskim tekstovima na srpskom jeziku
Automatic recognition and normalization of temporal expressions in Serbian unstructured newspaper and medical texts
Author
Jaćimović, Jelena B.
Mentor
Krstev, Cvetana
Committee members
Utvić, Miloš
Dragićević, Rajna

Grga, Đurica

Vitas, Duško

Metadata
Show full item recordAbstract
Ljudi u svakodnevnom životu koriste vreme kao univerzalni referentni sistem,
u okviru koga se doga¯daji ili stanja nižu jedan za drugim, utvr¯duje dužina njihovog
trajanja i navodi kada se neki doga¯daj desio. Znaˇcenje vremena i naˇcin
na koji ˇcovek poima vreme ogledaju se i u komunikaciji, pre svega, u jeziˇckim
izrazima koji se uˇcestalo koriste u svakodnevnom govoru. Vremenski izrazi, kao
fraze prirodnog jezika koje na direktan naˇcin ukazuju na vreme, pružaju informaciju
o tome kada se nešto dogodilo, koliko dugo je trajalo ili koliko ˇcesto se dešava.
Uporedo s razvojem informatiˇckog društva, pove´cava se i koliˇcina slobodno dostupnih
digitalnih informacija, što daje ve´ce mogu´cnosti pronalaženja potrebnih
informacija, ali i utiˇce na složenost ovog procesa, iziskuju´ci koriš´cenje naprednih
raˇcunarskih alata i mo´cnijih metoda automatske obrade tekstova prirodnih jezika.
S obzirom na to da se znaˇcenje ve´cine elektronskih informacija menja u zavisnosti
od vremena iskazanog... u njima, radi uspešnog razumevanja tekstova pisanih
prirodnim jezikom, neophodno je koriš´cenje alata koji su sposobni da automatski
oznaˇce i informacije koje referišu na vreme i omogu´ce uspostavljanje hronološkog
sleda opisanih doga¯daja. Stoga je potrebno razviti alate namenjene ekstrakciji
vremenskih izraza, kod kojih su preciznost i odziv na visokom nivou i koji se brzo
i jednostavno mogu prilagoditi novim zahtevima ili tekstovima drugog domena.
Postojanje ovakvog sistema može u velikoj meri uticati na poboljšanje uˇcinka primene
mnogih drugih aplikacija iz oblasti jeziˇckih tehnologija (ekstrakcija informacija,
pronalaženje informacija, odgovaranje na pitanja, rezimiranje teksta itd.), ali
i doprineti oˇcuvanju srpskog jezika u savremenom digitalnom okruženju...
People in everyday life use time as a universal reference system, within which,
events or states are sequenced one after the other, it is established how long they
lasted and it is stated when an event occurred. The meaning of time and the
way humans perceive time is reflected in communication, most of all, in linguistic
expressions frequently used in everyday speech. Temporal expressions, as natural
language phrases which directly refer to time, provide information on when something
happened, how long it lasted and how often it occurs.
Alongside with the information society development, the amount of freely available
digital information has increased, which provides a greater possibility of
finding the necessary information, but also affects the complexity of this process,
by requiring the use of advanced computer tools and more powerful natural language
text processing methods. Having in mind that the meaning of most electronic
information can change depending on time expressed in th...em, it is essential to
use tools which can both automatically mark the information related to time and
enable the establishment of chronological order of described events. Therefore,
it is necessary to develop tools for extraction of temporal expressions with high
levels of precision and recall, which can be easily and quickly adapted to new demands
and texts from different domains. The existence of such a system can, to
a great extent, affect the effectiveness improvement in implementation of many
other applications from the field of language technology (information extraction,
information retrieval, question answering, text summarization, etc.), but also contribute
to the preservation of the Serbian language in the contemporary digital
environment...