Linking the Thoughts: Analysis of Argumentation Structures in Scientific Publications

27351

Authors

nan

Abstract

This paper presents the results of an annota- tion study focused on the fine-grained analysis of argumentation structures in scientific pub- lications. Our new annotation scheme spec- ifies four types of binary argumentative rela- tions between sentences, resulting in the rep- resentation of arguments as small graph struc- tures. We developed an annotation tool that supports the annotation of such graphs and carried out an annotation study with four an- notators on 24 scientific articles from the do- main of educational research. For calculating the inter-annotator agreement, we adapted ex- isting measures and developed a novel graph- based agreement measure which reflects the semantic similarity of different annotation graphs. 1

Keywords

nan

Full Content

Proceedings of the 2nd Workshop on Argumentation Mining, pages 1–11, Denver, Colorado, June 4, 2015. c⃝2015 Association for Computational Linguistics Linking the Thoughts: Analysis of Argumentation Structures in Scientific Publications Christian Kirschner⋄†, Judith Eckle-Kohler⋄, Iryna Gurevych⋄† ⋄ UKP Lab, Technische Universit¨at Darmstadt † German Institute for Educational Research http://www.ukp.tu-darmstadt.de Abstract This paper presents the results of an annota- tion study focused on the fine-grained analysis of argumentation structures in scientific pub- lications. Our new annotation scheme spec- ifies four types of binary argumentative rela- tions between sentences, resulting in the rep- resentation of arguments as small graph struc- tures. We developed an annotation tool that supports the annotation of such graphs and carried out an annotation study with four an- notators on 24 scientific articles from the do- main of educational research. For calculating the inter-annotator agreement, we adapted ex- isting measures and developed a novel graph- based agreement measure which reflects the semantic similarity of different annotation graphs. 1 Introduction Argumentation mining aims at automatically identi- fying arguments and argumentative relations in ar- gumentative discourse, e.g., in newspaper articles (Feng and Hirst, 2011; Florou et al., 2013), legal documents (Mochales-Palau and Moens, 2011), or scientific publications. Many applications, such as text summarization, information retrieval, or faceted search could benefit from a fine-grained analysis of the argumentation structure, making the reasoning process directly visible. Such an enhanced infor- mation access would be particularly important for scientific publications, where the rapidly increas- ing amount of documents available in digital form makes it more and more difficult for users to find specific information nuggets without investing a lot of time in reading (parts of) documents which are not relevant. According to well-established argumentation the- ories in Philosophy and Logic (e.g. Toulmin (1958), Freeman (2011), Walton et al. (2008)), an argument consists of several argument components which of- ten are of a specific type, such as premise or claim. Argumentative relations are usually directed rela- tions between two argument components. Different relation types are distinguished, like support or at- tack (Peldszus and Stede, 2013) which indicate that the source argument component is a reason or a refu- tation for the target component. Argument compo- nents and argumentative relations together form the argumentation structure. Figure 1 shows the argu- mentation structure of one argument consisting of 6 argument components and 6 relations between them. Previous work has developed approaches to classify 1 We observe worse performance for children with migration background in primary school. In our study their school grade was significant worse. However there are high differences between the children. Children with language difficulties most often have to repeat a year. By contrast, children with a positive socio-economic background usually don't have any problems. supports attacks sequence supports supports 3 5 4 2 By socio-economic background we mean for example the parent's education level. 6 details Figure 1: Illustration of one argument consisting of 6 ar- gument components. 1 sentences in scientific papers according to their ar- gumentative role (Teufel, 1999; Liakata et …

Content truncated to 500 words

Labeled Text

This paper presents the results of an annotation study focused on the fine-grained analysis of argumentation structures in scientific publications. Our new annotation scheme specifies four types of binary argumentative relations between sentences, resulting in the representation of arguments as small graph structures. We developed an annotation tool that supports the annotation of such graphs and carried out an annotation study with four annotators on 24 scientific articles from the domain of educational research. For calculating the inter-annotator agreement, we adapted existing measures and developed a novel graphbased agreement measure which reflects the semantic similarity of different annotation graphs. 1

Publication Details

Paper ID:
27351
Year:
2015
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan