Collaborative Annotation and Visualization of Functional and Discourse Structures

Collaborative Annotation and Visualization of Functional and Discourse Structures Hengbin Yan Halliday Centre for Intelligent Applications of Language Studies, Department of Chinese, Translation and Linguistics, City University of Hong Kong hbyan2@cityu.edu.hk Jonathan Webster Halliday Centre for Intelligent Applications of Language Studies, Department of Chinese, Translation and Linguistics, City University of Hong Kong ctjjw@cityu.edu.hk Abstract Linguistic annotation is the process of adding additional notations to raw linguistic data for descriptive or analytical purposes. In the tagging of complex Chinese and multilingual linguistic data with a sophisticated linguistic framework, immediate visualization of the complex multi-layered functional and discourse structures is crucial for both speeding up the tagging process and reducing errors. The need for large-scale linguistically annotated corpora has made collaborative annotation increasingly essential, and existing annotation tools are inadequate to the task of providing assistance to annotators when dealing with complex linguistic structural information. In this paper we describe the design and development of a collaborative tool to extend existing annotation tools. The tool improves annotation efficiency and addresses certain difficulties in representing complex linguistic relations. Here, we adopt annotation based on Systemic Functional Linguistics and Rhetorical Structure Theory to demonstrate the effectiveness of the interface built on such infrastructure. Keywords: Linguistic Annotation, Linguistic Visualization, Cross-domain References 1. Introduction Recent years have witnessed an increasing need for large-scale high-quality annotated corpora on complex Chinese linguistic information where no automated annotators are available. Annotation on multi-level data complex structural relationships in such linguistic frameworks as Systemic Functional Grammar (SFG) [1] and Rhetorical Structure Theory [2] is a difficult task. SFG investigates texts as intentional acts of meaning, organized in functional-semantic components known as “metafunctions”. Three primary metafunctions, operating in parallel and each representing a layer of meaning with a set of options to language users, cover different functional aspects of human communication and expression: the ideational, Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012) 366 interpersonal and textual metafunctions. For our purposes our discussion will focus on analysis and annotation of these three metafunctions in SFG. Despite the fact that SFG is becoming increasingly influential among Chinese linguistic researchers, a large-scale, high-quality corpus annotated with SFG has yet to be developed [3]. Consequently, when trying to conduct corpus-based analysis using the SFG framework researchers must either 1) spend an enormous amount of time studying an unannotated corpus, 2) embark on the error-prone process of manually annotating a corpus on their own, or 3) rely on small corpora independently annotated by researchers which may not be particularly suited to needs of the tasks at hand. The lack of high-quality Chinese SFG corpora is partly attributable to the lack of a competent SFG tagger capable of annotating large-scale corpora while ensuring quality. In developing such a tagger, a number of challenges need to be addressed: 1) Lack of an efficient and sophisticated storage scheme for storing such multilayered information with complex structures 2) Additional visual cues to facilitate the tagging process 3) Need for collaborative tasking (co-tasking) by different …

Content truncated to 500 words