Abstract: A technique for processing natural language text uses a data structure that includes structure data in the text data. The structure data indicates an autonomous punctuational structure of the text, a punctuational structure that is independent of the lexical content of the text and therefore can be manipulated without considering the meaning of the words in the text. The data structure can be a tree in which each node has a textual type such as a paragraph, sentence, clause, phrase, or word. The data structure could alternatively be parallel data sequences, one with codes indicating the text's characters and the other with codes indicating textual types. The data structure is produced and maintained using a grammar of textual types, indicating for each textual type the textual types of units into which it can properly be divided. During editing, a text sequence is generated by applying rendering rules to the data structure, and the text is presented to the user based on the text sequence.