Search

In multisentence texts, the order and interrelationships of sentence topics is of crucial importance if the reader is to understand easily. What makes paragraphs coherent? What strategies do people employ to control the presentation order and linking of material? Without a theory of coherence, text generation systems have little hope of producing acceptable texts. While various theories of text coherence have been developed, no single theory captures all the phenomena of human-generated paragraphs. In this paper we argue that the coherence of a paragraph does not result from the application of a single theory, but instead results from the cooperation of a number of different coherence strategies. W e illustrate this claim by showing how two very different theories about the planning of coherent text — 1) Rhetorical Structure Theory, based on structural and semantic relationships that hold between pieces of the text, and 2) Focus Trees, based on how the focus of attention shifts during discourse — can be used within a single system to complement each other to best advantage.

How is affect communicated in language? Natural languages contain a large number oftechniques for injecting affect into text, both explicitly and implicitly. This paper discussessome techniques that speakers use to slant their text, and describes what is required for acomputer program to generate differently slanted versions of a single underlying representation.

This paper addresses part of the question "how do we say the same thing in different ways in order to communicate non-literal, pragmatic information?". Since the style of the text can communicate much information — it may be stuffy, slangy, prissy — generators that seek to satisfy pragmatic, hearer-related goals in addition to simple informative ones must have rules that control how and when different styles are used. But what is "style" ? In this paper,formal and informal language is analysed to provide stylistic rules that enable a program to produce texts of various levels of formahty

This report summarizes the research and development work done over four years toward the goal of automatically planning and generating fluent multisentence paragraphs of English text. The work consisted of three principal components, namely knowledge representation, grammar development, and text structuring. With respect to knowledge representation, a powerful technique of linking the generator with arbitrary applications was developed by using a very general underlying taxonomy of entities in the world and various specific domain-related taxonomies. As part of grammar development, the invertibility of the grammar in use by the project was investigated, with the eventual goal of developing a combined bidirectional parsing-generation system using the same grammar net-work. Finally, a text structure planner was developed and the whole system was successfully used to generate paragraphs in three different application domains.

This paper discusses the planning of multisententialal text by computer. In order to construct coherent paragraphs, we have been using relations from Rhetorical Structure Theory (RST) operationalized as plans. The paper first describes, in some detail, the current method of planning a paragraph using operationalized RST relation/plans. It then makes two points that illustrate why RST relation/plans are the ideal tool for planning paragraphs. First, these relation/plans can be shown to combine the best features of paragraph-sized schemas and clause-sized planning rules under a top- down planning regime in a way which affords much flexibility to the user. Second, RST relation/plans can support both standard top-down planning and open- ended conversation-like behavior; a small difference in treatment gives rise to either paradigm. Keywords: Coherence, Computational linguistics, Discourse, Natural language generation, Natural language processing, Penman, Rhetorical structure theory, Text planning., Presented at the International Workshop on Text Generation (4th), Catalina Island, CA, Jul 1988.

Natural Language generation programs written to date are not sensitive to anything but their input, and therefore produce the same output to all hearers in all circumstances. For a generator to produce the various versions of a single input it must have the ability to make appropriate choices about the content and form of its text. Due to the flexibility of language, speakers can communicate far more than just the literal content of the words they use; the additional information usually serves some of their audience-related goals. Thus the generator's choices should be governed by the goals resulting from pragmatic issues such as its knowledge of the hearer and the setting of the conversation. This research investigates the types of interpersonal goals that speakers can have; how they can be made specific enough to direct the generation process; the interpretation of input, but generator-directed inference, to find suitable forms of expression; and the representation of language in a phrasal lexicon. In the model of generation that incorporates these goals, planning and realization are interleaved processes, where the interleaving takes place at choice points. This view supports the standard top-down planning-to-realization approach as well as the bottom-up approach in which linguistic options present themselves as opportunities for the achievement of the active goals. To illustrate these ideas, the program PAULINE (Planning And Uttering Language In Natural Environments) produces, from a single set of input representations in each of three domains, various paragraphs that differ in slant, content, and style.