Towards Domain-Independent Argumentative Zoning: Evidence from Chemistry and Computational Linguistics

Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 1493–1502, Singapore, 6-7 August 2009. c⃝2009 ACL and AFNLP Towards Discipline-Independent Argumentative Zoning: Evidence from Chemistry and Computational Linguistics Simone Teufel Computer Laboratory Cambridge University sht25@cl.cam.ac.uk Advaith Siddharthan Computer Laboratory Cambridge University as372@cl.cam.ac.uk Colin Batchelor Royal Society of Chemistry Cambridge, UK batchelorc@rsc.org Abstract Argumentative Zoning (AZ) is an anal- ysis of the argumentative and rhetorical structure of a scientific paper. It has been shown to be reliably used by independent human coders, and has proven useful for various information access tasks. Annota- tion experiments have however so far been restricted to one discipline, computational linguistics (CL). Here, we present a more informative AZ scheme with 15 categories in place of the original 7, and show that it can be applied to the life sciences as well as to CL. We use a domain expert to encode basic knowledge about the sub- ject (such as terminology and domain spe- cific rules for individual categories) as part of the annotation guidelines. Our results show that non-expert human coders can then use these guidelines to reliably an- notate this scheme in two domains, chem- istry and computational linguistics. 1 Introduction Teufel et al. (1999) define the task of Argumenta- tive Zoning (AZ) as a sentence-by-sentence clas- sification with mutually exclusive categories from the annotation scheme given in Fig. 1. The reason- ing behind the categories is inspired by the notion of a knowledge claim (Myers, 1992; Luukkonen, 1992): the act of writing a paper corresponds to an attempt of claiming ownership for a new piece of knowledge, which is to be integrated into the repository of scientific knowledge in the authors’ field by the process of peer review and publica- tion. In the cause of this process, the authors have to convince the reviewers that the knowledge claim of the paper is valid (Swales, 1990; Hy- land, 1998). What AZ aims to model, then, are some of the relevant stages in this argument. We divide the paper into zones, OTHER, OWN and BACKGROUND. These are defined on the basis of who owns the knowledge claim in the corre- sponding segment. There are also two categories which are defined by their relationship to existing work, BASIS and CONTRAST. That means that parts of the AZ scheme are similar to citation func- tion classification schemes from the area of cita- tion content analysis (Garfield, 1965; Weinstock, 1971; Spiegel-R¨using, 1977), and to automatic citation function classification (Nanba and Oku- mura, 1999; Garzone and Mercer, 2000; Teufel et al., 2006). The remaining categories, AIM and TEXTUAL, fulfil different rhetorical functions for the presentation of the paper. AIM points out the paper’s main knowledge claim, a rhetorical move which may be repeated in the conclusion and the introduction. TEXTUAL explains the physical lo- cation of information, e.g., by giving a section overview or presenting a summary of a subsec- tion. On the basis of human-annotated training material, AZ can be automatically classified using supervised machine …

Content truncated to 500 words