A Text Plan Method Based On XML

Text generation is an important target of natural language generation research. And text plan is an indispensable part of text generation. Text plan is used to determine and structure the produced content mainly, plays a key role to ensuring the produced content's quality. The current method of text plan should be perfected by reducing the complexity of implementation, reflecting user's intention and increasing the variety of text structure produced. The method of text plan based on XML, which is propounded by this dissertation, just march towards such direction showing as above. On the basis of summarizing the relevantly work both at home and abroad and studying the marrows of Schema and RST, which are two mainstream methods of text plan, this dissertation propounds a text plan method based on XML, named KB(knowledge base)-Information-Plan. It is a layered strategy of plan. The high level uses the marrow of RST theory to plan text's general structure in a flexible way, namely divides the sub text span (the whole text is the biggest text span), determines the divided sections's discourse relation and arranges their order. The bottom level uses the marrow of Schema method to plan every sub text span fast. The dissertation provides the method's implementation model, defines and constructs the relevant components: XML-info-tree and XML-plan-tree. The former represents user information and communion information using frame. The latter records the structure information, namely plan's result. In order to alleviate the difficulty of realizing RST theory, we introduce XML language, and propose an idea for the first time. The idea is using frame to represent communion information in XML and constructing object oriented KB based on special rules,named XML-KB. The method is fit for such text generation system, which draws information from the existing knowledge base, according to a certain confirmed goal. And, the method covers the advantages offered by Schema and RST.