Pagrindinis šio darbo tikslas - išanalizuoti lietuvių kalbos semantinio anotavimo procesą ir sukurti metodologiją, leisiančią įgyvendinti semantinio anotavimo proceso automatizavimą. Kompiuteris, kitaip nei žmogus, nesupranta ar tekstas yra rišlus ir prasmingas, ar neturi jokio rišlumo ir prasmės. Tai ypač atsiliepia verčiant tekstus, lyginant tos pačios kalbos tekstus vieną su kitu ir t.t. Semantinis anotavimo procesas leidžia išspręsti šią problemą sukurdamas metodiką, kuri leidžia aprašyti žodžių sąsajas sakiniuose, tačiau norint tekstą semantiškai suanotuoti rankiniu būdu, reikia turėti specifinių žinių ir tai reikalauja didelių laiko sąnaudų. Norint išvengti šių problemų semantinį procesą būtina automatizuoti. Šiame darbe buvo išanalizuotas lietuvių kalbai tinkantis semantinio anotavimo procesas, sukurta metodologija, leidžianti įgyvendinti semantinio anotavimo proceso automatizavimą. Metodologijos pagrindu JAVA programavimo kalba įgyvendintas automatizuoto semantinio anotavimo proceso realizavimas, atliktas eksperimentas ir pateiktos išvados. / Major aim of this work – analyze Lithuanian language semantic annotation process and develop methodology which let implement automate semantic annotation process. The computer, unlike the person, does not understand did the text is coherent and meaningful, or have no coherence and meaning. This is particularly vulnerable to the translation of the text, compared with the same language texts with one another, etc. Semantic annotation process allows to solve the problem of creating a methodology that enables to describe links between words in sentences, but in order to semantically annotate the text manually, you need to have specific knowledge and it requires time-consuming. To avoid these problems it is necessary to automate the process of semantic. In this work was analyzed the Lithuanian language suitability for semantic annotation process, developed methodology to implement semantic annotation of process automation. Methodology based Java programming language implementation of automated semantic annotation process realization of the experiment was conducted and the following conclusions.

<p>Nowadays, Dynamic Geometry Software (DGS) is widely accepted as a tool for creating and presenting visually rich interactive teaching and learning materials, called dynamic drawings. Dynamic drawings are specified by writing expressions in functional domain-specific languages. Due to wide acceptance of DGS, there has arisen a need for their extensibility, by adding new semantics and visual objects (visuals). We have developed a programming framework for the Dynamic Geometry Software, SLGeometry, with a genericized functional language and corresponding expression evaluator that act as a framework into which specific semantics is embedded in the form of code annotated with metadata. The framework transforms an ordinary expression tree evaluator into an object-oriented one, and provide guidelines and examples for creation of interactive objects with dynamic properties, which participate in evaluation optimization at run-time. Whereas other DGS are based on purely functional expression evaluators, our solution has advantages of being more general, easy to implement, and providing a natural way of specifying object properties in the user interface, minimizing typing and syntax errors.LGeometry is implemented in C# on the .NET Framework. Although attributes are a preferred mechanism to provide association of declarative information with C# code, they have certain restrictions which limit their application to representing complex structured metadata. By developing a metadata infrastructure which is independent of attributes, we were able to overcome these limitations. Our solution, presented in this&nbsp; dissertation, provides extensibility to simple and complex data types, unary and binary operations, type conversions, functions and visuals, thus enabling developers to seamlessly add new features to SLGeometry by implementing them as C# classes annotated with metadata. It also provides insight into the way a domain specific functional language of dynamic geometry software can be genericized and customized for specific needs by extending or restricting the set of types, operations, type conversions, functions and visuals.Furthermore, we have conducted&nbsp; experiments with several groups of students of mathematics and high school pupils, in order to test how our approach compares to the existing practice. The experimental subjects tested mathematical games using interactive visual controls (UI controls) and sequential behavior controllers. Finally, we present a new evaluation algorithm, which was compared to the usual approach employed in DGS and found to perform well, introducing advantages while maintaining the same level of performance.</p> / <p>U dana&scaron;nje vreme softver za dinamičku geometriju (DGS) je &scaron;iroko prihvaćen kao alat za kreiranje i prezentovanje vizuelno bogatih interaktivnih nastavnih materijala i materijala za samostalno učenje, nazvanih dinamičkim crtežima. Kako je raslo prihvatanje softvera za dinamičku geometriju, tako je i rasla potreba da se oni pro&scaron;iruju, dodajući im novu semantiku i vizualne objekte. Razvili smo programsko okruženje za softver za dinamičku geometriju, SLGeometry, sa generičkim&nbsp; funkcionalnim jezikom i odgovarajućim evaluatorom izraza koji čini okruženje u kom su ugrađene specifične semantike u obliku koda označenog metapodacima. Ovo okruženje pretvara uobičajen evaluator stabla izraza u objektno orijentiran, te daje uputstva i primere za stvaranje interaktivnih objekata sa dinamičkim osobinama, koji sudeluju u optimizaciji izvr&scaron;enja tokom izvođenja. Dok se drugi DGS-ovi temelje na čisto funkcionalnim evaluatorima izraza, na&scaron;e rje&scaron;enje ima prednosti jer je uop&scaron;tenije, lako za implementaciju i pruža prirodan način navođenja osobina objekta u korisničkom interfejsu, minimizirajući kucanje i sintaksne gre&scaron;ke. SLGeometry je implementirana u jeziku C# .NET Framework-a. Iako su atributi preferiran mehanizam, koji povezuje C# k&ocirc;d sa deklarativnim informacijama, oni imaju određena ograničenja koja limitiraju njihovu primenu za predstavljanje složenih strukturiranih metapodataka. Razvijanjem infrastrukture metapodataka koja je nezavisna od atributa, uspeli smo prevladati ta ograničenja. Na&scaron;e re&scaron;enje, predstavljeno u ovoj disertaciji, pruža pro&scaron;irivost: jednostavnim i složenim vrstama podataka, unarnim i binarnim operacijama, konverzijama tipova, funkcijama i vizuelnim objektima, omogućavajući&nbsp; time programerima da neprimetno dodaju nove osobine u SLGeometry&nbsp; implementirajući ih kao C# klase označene metapodacima.</p>