DiZer 2.0 – a Web Interface for Discourse Parsing Erick Galani Maziero, Thiago Alexandre Slagueiro Pardo Núcleo Interinstitucional de Lingüística Computacional (NILC) Instituto de Ciências Matemáticas de Computação, Universidade de São Paulo Av. Trabalhador São-carlense, 400. P.O.Box. 668. 13560-970 - São Carlos/SP, Brazil {erickgm,taspardo}@icmc.usp.br Abstract. This paper presents the DiZer 2.0, a freely available web interface for discourse parsing. Different from its first version, it is easier to use and allows easy customization for other text genres, domains and languages. Keywords: discourse parsing, RST 1 Introduction Discourse parsing aims at automatically identifying the discourse structure of a text. RST (Rhetorical Structure Theory) [1] has been the most followed discourse model. According to RST, a text may be structured as a sequence of segments/Elementary Discourse Units (EDUs) (usually clauses or sentences) connected by rhetorical relations, forming a tree-like structure. See, e.g., Fig. 1 below, where each segment is numbered. N indicates the segments that are nuclear and, therefore, are more important in the text; S, otherwise, indicates satellite segments, which are considered complementary information. Fig. 1. Example of rhetorical structure Some discourse parsers exist for English, Portuguese, Spanish, and Japanese, to the best of our knowledge. DiZer [2] is the only known parser for Portuguese. DiZer was originally designed to academic texts and uses a set of discourse patterns (that correlate text markers and discourse relations) to identify text structure. Its main drawbacks are that it is a heavy system (requiring the installation of several tools and resources) and is hard to adapt and port to any other text genre, domain, and language. In order to deal with this, we proposed DiZer 2.0, a web interface that (i) do not require any tool or resource to be locally installed, (ii) is easier to use, and (iii) may be easily adapted to other texts and languages. DiZer 2.0 is introduced in the next section.