Emina Kurtic

Department of Human Communication Sciences
The University of Sheffield
31 Claremont Crescent
Sheffield
S10 2TA
UK

Department of Computer Science
The University of Sheffield
Regent Court
211 Portobello Street
Sheffield
S1 4DP
UK

email : emina@dcs.shef.ac.uk

Biography

I am a PhD student at the Department of Human Communication Sciences and the Department of Computer Science at the University of Sheffield. In my PhD project on overlapping speech in meetings I investigate how a phenomenon commonly found in spontaneous multi-party speech can be studied on large data sets using statistical modelling methods. My project is supervised by Prof Bill Wells and Dr Guy Brown.

Before coming to Sheffield, I obtained a Magister Artium (MA) degree in Linguistics from the Ruhr-Universität Bochum, Germany. I also hold an MSc in Speech and Language Processing from the University of Edinburgh.

Recent Projects

Overlapping talk in spontaneous multi-party conversations

PhD project funded by a University Project Studentship of the University of Sheffield

Overlapping speech is very common in spontaneous conversations although it is not a default mode of communication. Previous work has acknowledged that overlapping speech can be employed by conversation parties as a means of turn competition. Alternatively, it can be noncompetitive and signal interactional collaboration between participants, or simply be a byproduct of regular turn-taking mechanisms. The main goal of this project is to investigate the differences in the linguistic design that is associated with these overlap functions.

The methodological approach adopted in this work combines conversation analysis (CA) with statistical modelling of overlapping talk. We draw upon previous findings on overlap-relevant linguistic cues identified within the CA framework in order to compile a large set of overlap relevant features. Unlike CA work, the conclusions about relevance of linguistic cues for overlap competitiveness are drawn based on results of statistical overlap models rather then on impressionistic analysis of a small data set. This methodology enables us to ground the conclusions on a representative data set employing objective measures. At the same time we keep the necessary level of detail which can only be obtained from CA. In this respect our work also departs from other data driven overlap studies, which are not able to relate their findings to the conversational function of the overlap.

Anaphora resolution in spontaneous spoken discourse

Recently, I also worked on anaphora resolution in spontaneous multi-party conversations. I applied Maximum Entropy to resolution of pronominal anaphora in ICSI meetings data.

Collaborators

Supervisors: Professor Bill Wells and Dr.Guy Brown