Theory and Method of Operation Document’s Automatic Generation
|School||Nanjing University of Technology and Engineering|
|Keywords||Operation Document NLG Content Planning Sentence Planning Surface Realization|
Making Operation Document is an important content in C3I system. At present, the ’Template-based’ method is used popularly. The NLG theory has provided nicer concept base for the automatic generation of Operation Document. So, this dissertation focuses on the theory, techniques and method of Operation Document’s automatic generation, based on NLG. They have great significance of theory and application to the intelligent generation of Operation Document.The main contributions of the dissertation are as follows:1. A new method of Military Document’s generation is presented.it analyzes the Operation Document’s characteristics,types of sentence.actuality.discusses the corpus analysis which can acquire knowledge for Operation Document’s Generation.establishes the Operation Document language mathematic modeling.2. A Multi-View architecture is presented. It contains Function-View,System-View and Technology-View, provides a general standard method for the development of NLG systems.lt describes the architecture fully.3. The semantic, syntax and lexical knowledge representation model are built. The Schema, which described the text structure of Operation Document, is realized by ATN. The rules in the Sentence Aggregation are described in the production way. Sentence Structure Classes and Resource Trees are adopted to depict syntax structure of sentences, which have similar semanteme. Complex Feature Set is used to describe dictionary.4. A new content planning model.which is in the automatic generation of Operation Document.combined with the Schema and RST approach is presented.In the combined model.Schema determines the global organization of the text, while RST controls the local coherence of the discourse.5. The Sentence Aggregation model and Language Resource Process model are established. The former makes and realizes the aggregation rules with the representation of Schema and RST.The later determines the place of semantic components,which identify the content.in the sentence.6. The FUG model,SFG model and statistical probability model are set up.FUG model uses the complex feature set.SFG model stresses the selection of character net.while probability model bases on corpus.