LiveTree: An Integrated Workbench for Discourse Processing

LiveTree: An Integrated Workbench for Discourse Processing Gian Lorenzo Thione, Martin van den Berg, Chris Culy, Livia Polanyi FX Palo Alto Laboratory 3400 Hillview Ave, Bldg. 4 Palo Alto, CA 94304 {thione|vdberg|culy|polanyi}@fxpal.com Abstract In this paper, we introduce LiveTree, a core component of LIDAS, the Lin- guistic Discourse Analysis System for automatic discourse parsing with the Unified Linguistic Discourse Model (U-LDM) (X et al, 2004). LiveTree is an integrated workbench for super- vised and unsupervised creation, stor- age and manipulation of the discourse structure of text documents under the U-LDM. The LiveTree environment provides tools for manual and auto- matic U-LDM segmentation and dis- course parsing. Document manage- ment, grammar testing, manipulation of discourse structures and creation and editing of discourse relations are also supported. 1 Introduction In this paper, we introduce LiveTree, a core component of LIDAS (the Linguistic Discourse Analysis System) for automatic discourse parsing with the Unified Lin- guistic Discourse Model (U-LDM) (Po- lanyi et al, 2004). The U-LDM is a theory of discourse structure and semantics that has as its goal assigning the correct inter- pretation to natural language utterances. 1.1 Overview of LiveTree LiveTree is an integrated workbench for supervised and unsupervised creation, storage and manipulation of the discourse structure of text documents under the U- LDM. LiveTree does not support speech, dialog or interaction annotation (Bernsen et al. 2002, 2003 and over view of systems in Bernsen et al. 2002). The LiveTree en- vironment provides tools for manual and automatic U-LDM segmentation and dis- course parsing. Like RSTTool, LiveTree provides support for segmentation, mark- ing structural relations among segments, and creating and editing discourse rela- tions (O’Donnell 2003). Similar to the D- LTAG system described in Forbes et al (2003) LiveTree is an experimental dis- course parser implementing a theory of sentential and discourse relations. How- ever, LiveTree is also a complete docu- ment handling and manual and automatic discourse parsing system. Various applica- tions are supported as web services. Ac- cordingly, LiveTree serves as both the user interface and theory development envi- ronment for PALSUMM, a text summari- zation system built on top of LIDAS (See Section 5 below) In this paper, we de- scribe the resources LiveTree workbench provides for discourse level theoretical development as well as document han- dling, manual and automatic text annota- tion and parsing. 1.2 LiveTree Functionalities LiveTree’s Java architecture shown in Figures 1 is modular and highly extensi- ble. LiveTree is made up by: (1) a Model Manager which provides interfaces for manipulation, storage and retrieval of ac- tual documents and discourse representa- tions; (2) a Module Manager, which han- dles and provides access to the main GUI and to all installed modules; and (3) a Ser- vice Manager providing a polling interface for all active LiveTree Services. Manager components rely on stubs which can be implemented and extended from outside the framework’s core. The LiveTree Module Manager and all installed LiveTree Modules lie on top of a general GUI Layer handling the main Li- veTree window, …

Content truncated to 500 words