Text or document planning as the mechanism of ordering messages in a coherent way for achieving a cohesive text has traditionally been realized by schemas or the establishment of rhetorical relations between messages and message sequences. Inspired by the descriptions of a comprehensive set of rhetorical relations in Rhetorical Structure Theory (RST), these relations have often been realized as planning operators for achieving a complete text. The rise of machine learning approaches for NLG tasks seems to hide the fact that RST-oriented text planning is by far not the only method for achieving linguistically relevant text plans.
Formal semantics and pragmatics offer a number of different theories on text organization and coherence phenomena whose explanatory power goes beyond the justified grouping of informational units, among them (Segmented) Discourse Representation Theory (SDRT), Question-under-discussion (QUD) approaches, and probabilistic approaches to meaning. This workshop addresses the relevance of these theories for text planning.
For example, QUD approaches to text structuring provide expressive theories facilitating concise analyses of a group of different pragmatic phenomena, ranging from the analysis of focus/background structures to dialogue moves, but they did not receive much attention for text planning issues in NLG. QUDs are the central concept in analyses that explain linguistic regularities as a consequence of the assumption that the sentences and text segments with which the regularities are associated are answers to an explicitly or implicitly asked question. QUDs were early on used for explaining possible sequences of dialogue moves (Carlson, 1983; Ginzburg, 1996), clarifying information structural concepts (e.g. the topic/focus distinction), temporal progression and foreground/background relations in narration, information structural constraints on implicatures (van Kuppevelt, 1996), representing discourse goals and defining contextual relevance (Roberts, 1996), and for analysing structure and coherence of discourse, of both text and dialogue (van Kuppevelt, 1995). Since then, QUDs have been firmly established as an analytic tool, leading to fruitful applications for a wide range of linguistic phenomena.
The aim of this workshop is to explore the interplay of linguistic theories of text planning-related phenomena with computational approaches to text planning, be it rule-based or learning approaches, in order to bring these fields of expertise together.
We invite submissions of extended abstracts (3 pages including references), related to all aspects of text planning from a linguistic, theoretical and computational point of view. Topics of the workshop include, but are not limited to, the following issues:
- computational models of text planning
- corpus annotation of text structures and its interaction with information structure, discourse relations, dialogue goals etc.
- formal approaches to pragmatics and their relation to issues of text planning
Papers should be submitted electronically through the START conference management system (to be opened in due course).
In order to foster the exchange of ideas, we intend a kind of semi-shared task for this workshop: We will provide a small corpus of annotated texts to be generated and a paper that addresses the workshop task so that each workshop speaker does not only present his/her own research, but outlines his or her own ideas on generating these texts. The analysis of one and the same texts from different theoretical perspectives should shed light on the pros and cons of using various theory-driven approaches to text planning. We will make the data available after the submission of the abstracts.
The workshop shall be of interest for linguists who are working at the interface of pragmatics and computational linguistics and people in the NLG community with an interest in linguistic theories of text organization.
important dates:
submissions due: September, 20, 2020
Notification of acceptance: October, 20, 2020
Camera ready papers due: November, 20, 2020