Automatic Spatial Layout Of Named Entity In Natural Sentence

Automatic Layout Method of Natural Language Described Spatial Entity is across-application of natural language processing and graphics computing. Itachieved the goal that the natural language automatically transformed into athree-dimensional graphics. You can get a three-dimensional scene without learnigcomplex software. Text to a three-dimensional scene has been applied to manyfields. For example, a traffic accident simulation, control agent in a virtual world. Itwill be applied in the field of education in the future and become a practical tool inthe aspect of communication.Currently, it has several difficulties in the process of text to three-dimensionalscene. The same word has different meanings in different contexts because of therichness of natural language. Therefore we need to combine with the context andbuild a knowledge base. For maximum effect, we need a large model library tomake the scene closer to reality. To adapt the models is also a big job, we shouldresize the models and mark on them. It may have conflict when we place the spatialentity, so we need a rule library to realize a optimized layout of spatial entities.The thesis includes following three contents mainly:1. Identify locality based on pattern matching. In this section, we build alocality recognition dictionary firstly. The dictionary contains all the simplelocalizers including the meaning item,usage and example sentences. On the basis ofthe dictionary,we analyze the usage of localizers and then build a pattern base toidentify localizers. We also improve the locality recognition dictionary combinedwith the locality corpus. After the experiment on the real corpus, we contrast theresult between automatic and manual tagging. We adjust the order of patterns andmodified some of them.2. Identify trajector and landmark based on dependencies. Dependencies showsthe relationship between words in a sentence, and analyzing the dependencies weare interested can get the information of color and size about the spatial entity. Thenthese dependencies are expressed to binary form group. We have studied how toidentify trajector and landmark base on dependencies. The trajector and landmarkhave been divided into two kinds, simple and complex. We give different algorithmto different kind. Finally, the information about trajector and landmark is writteninto a file which we call it described file.3. Automatically generate scene based on java3D. In this section we proposethe technology of automatically generating a scene base on Java3D. We described the process of the file which generated in the previous module to the scene. We alsobuild a model library. In order to meet the real word we adjust the size andorientation of the models. We propose a method to resolve conflict between thespatial entities by divide them into groups. To identify the implicit information inthe natural language, we build a rule library.