Teksta nozīmes vizualizēšana
Loading...
Date
Authors
Advisor
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Language
N/A
Abstract
Dabiskās valodas apstrāde ir vienkāršs un efektīvs veids, lai aprakstītu vizuālu scēnu.
Izprast semantiskas dabiskās valodas ir grūts un sarežģīts uzdevums datoriem. Šī darba
mērķis ir izpētīt esošās zināšanu attēlojuma metodes un izveidot automātisku sistēmu
animācijas scēnai veidošanai no teksta. FrameNet pieeja tiek izmantota vārdu nozīmes
attēlošanai un tā nodrošina zināšanu bāzei semantisko lomu marķēšanai. FrameNet ļauj
apvienot teksta vārdus semantiskajā nozīmi, pamatojoties uz tematisko lomu ideju kā uz
centrālo nozīmes aspekts. Teksta uz scēnu pārveides process sastāv no trim posmiem.
Informācijas ieguves posms rada semantiskā struktūras attēlojumu. Nākamais posms nosaka
vizuālus elementus un ainas darbības, izmantojot plānošanas sfēras definīcijas valodu.
Pēdējais posms rada scēnas animāciju.
Natural language is an easy and effective way for describing visual scene. Understanding natural language semantic is a difficult and complex task to computers. The goal of this work is to investigate existing knowledge representation techniques and present automatic language-based system for creating animated scene from text. FrameNet approach is used for mapping towards word-meanings and provides a knowledge base for semantic role labeling. FrameNet allows to combine words in the text by semantic meaning, based on the idea of thematic roles as a central aspect of meaning. The text-to-scene conversion process consists of three stages. An information extraction module creates a semantic structure representation. Next module determines visual elements and actions on the scene using planning domain definition language. Final module creates animation of the scene.
Natural language is an easy and effective way for describing visual scene. Understanding natural language semantic is a difficult and complex task to computers. The goal of this work is to investigate existing knowledge representation techniques and present automatic language-based system for creating animated scene from text. FrameNet approach is used for mapping towards word-meanings and provides a knowledge base for semantic role labeling. FrameNet allows to combine words in the text by semantic meaning, based on the idea of thematic roles as a central aspect of meaning. The text-to-scene conversion process consists of three stages. An information extraction module creates a semantic structure representation. Next module determines visual elements and actions on the scene using planning domain definition language. Final module creates animation of the scene.