Contributions to Shallow Discourse Parsing: To English and Beyond (M. Kurfali)

Murathan Kurfali is a PhD student in Computational Linguistics. On March 15, he will defend his dissertation Contributions to Shallow Discourse Parsing: To English and beyond.

Join the webinar in Zoom

PHD DEFENCE: MURATHAN KURFALI

Contributions to Shallow Discourse Parsing: To English and beyond

Opponent: Amir Seldes, Georgetown University
Supervisors: Robert Östling, Mats Wirén, Christian Hardmeier

Link to dissertation in DiVA

Murathan Kurfali’s profile page

Abstract

Simply put, discourse is a coherent set of sentences where the sequential reading of the sentences yields a sense of accumulation and readers can easily follow why one sentence follows another. A text that lacks coherence will most certainly fail to communicate its intended message and leave the reader puzzled as to why the sentences are presented together. However, formally accounting for the differences between a coherent and a non-coherent text still remains as a challenge. Various theories propose that the semantic links that are inferred between sentences/clauses, known as discourse relations, are the building blocks of the discourse which can be connected one another in various ways to form the discourse structure. This dissertation focuses on the former problem of discovering such discourse relations without aiming to arrive at any structure, a task known as shallow discourse parsing (SDP). Unfortunately, so far, SDP has been almost exclusively performed on the available gold annotations of discourse relations in English, leading only limited insight into how the existing models would perform in the low-resource scenario which applies to, arguably, any non-English language. The main objective of the current dissertation is to address this shortcoming and help extend SDP to the non-English territory. The goal is pursued through three different threads: (i) investigation towards what kind of supervision would be minimally required to perform SDP, (ii) construction of multilingual resources annotated at discourse-level, (iii) extension of well-known means to (SDP-wise) low-resource languages. Additionally, as the secondary aim, the feasibility of SDP as a probing task to evaluate discourse-level understanding abilities of modern language models is also explored.

The dissertation is based on six papers grouped under three themes. The first two papers perform different subtasks of SDP through relatively understudied means. Paper I presents a simplified end-to-end method to perform explicit discourse relation labeling without any feature-engineering whereas Paper II shows how implicit discourse relation recognition benefits from large amounts of unlabeled text through a novel method for distant supervision. Both papers aim to gain insight regarding what is essential or not (i.e. hand-crafted features, annotated examples) for SDP by challenging the conventional forms of supervision used in these tasks in the hope of guiding future research on low-resource languages. The third and fourth papers describe two novel multilingual discourse resources, TED-MDB (Paper III) and three bilingual discourse connective lexicons (Paper IV). Notably, Ted-MDB is the first parallel corpus annotated for PDTB-style discourse relations covering six non-English languages. The last two studies directly deal with multilingual discourse parsing. Paper V reports the first results in cross-lingual implicit discourse relation recognition using Ted-MDB corpus, showing that even in the zero-shot setting, cross-lingual classification of implicit discourse relations is possible. Paper VI, on the other hand, proposes a multilingual benchmark including certain discourse-level tasks that have not been explored in this context before. Through this benchmark, the modern multilingual language models are evaluated for their ability to model interactions beyond the sentence level to guide future work in SDP towards best performing models. In summary, the dissertation is intended to serve towards building high-performance multilingual (or non-English monolingual) shallow discourse parsers, through the proposed methodologies and the constructed resources.

Contact

Robert Östling: robert@ling.su.se

Opponent

Amir Zeldes, Georgetown University

Supervisors

Robert Östling, Department of Linguistics, Stockholm University
Mats Wirén, Department of Linguistics, Stockholm University
Christian Hardmeier, IT University of Copenhagen

Chairman

Ljuba Veselinova, Department of Linguistics

Grading committee

Sara Stymne, Department of Linguistics and Philology, Uppsala University
Richard Johansson, Computer Science and Engineering, University of Gothenburg and Chalmers University of Technology
Bernhard Wälchli, Department of Linguistics, Stockholm University
Elena Volodina, Department of Swedish, Multilingualism, Language Technology, University of Gothenburg (reserve member)