Discourse Processing for Social Media Texts

Course organization

Instructors:	Tatjana Scheffler Manfred Stede
Time:	Fridays, 10-12 a.m. summer semester 2015
Place:	Golm, building 5, room 004
Module:	CSAM11 or CSAM12
Moodle:	Please register on the course's Moodle site.

Requirements

regular readings
active participation
possibly a short presentation
project work + documentation (final paper)

Course description

A central factor of text coherence is the speaker’s/author’s establishing discourse relations between adjacent portions of text, e.g., temporal or causal ones. In this class, we will work with two influential theories of discourse structure that build on such relations:

• The Penn Discourse Tree Bank analyzes explicit and implicit connectives (e.g., because, nevertheless, afterwards) in text and links them to their arguments, i.e. the text spans that are joined by the relation.
• Rhetorical Structure Theory posits a complete tree structure to arise from assigning individual relations to adjacent segments and stipulates certain well-formedness constraints for this structure.

In the first phase of the class, we will study both approaches in some detail. Then we’ll select one (or both) and aim for an implementation of a discourse parser. Annotated data (in English and/or German) is available for both approaches. Importantly, we will apply the parsing task not only to "standard" newspaper text (as earlier work has done) but also to Social media data, especially microblogs. This is a new field of application that investigates the mechanics of conversations and their coherence, for example on Twitter.

The course grade will be based mainly on the project work in the second phase of the seminar, and a final report you will write about it.

Literature

(to be extended...)

• William Mann and Sandra Thompson. 1988. Rhetorical Structure Theory: Towards a Functional Theory of Text Organization. TEXT, 8:243–281.
• R. Prasad, N. Dinesh, A. Lee, E. Miltsakaki, L. Robaldo, A. Joshi and B. Webber. 2008. The Penn Discourse Treebank 2.0. In Proc. of the 6th International Confer- ence on Language Resources and Evaluation (LREC). Marrakech, Morocco.