Alessandra Zarcone - Thesis Topics

NLP

Modelling the influence of type and thematic fit in linguistic phenomena using compositional distributional semantics

This is a proposal for a Master's thesis in compositional distributional semantic modelling of psycholinguistic data. The work involves exploring compositional distributional semantic models such as the Practical Lexical Function Model (Paperno et al, 2014) to model an existing psycholinguistic dataset which manipulates both the influence of type and thematic fit. The model will have the potential to be extended to model a wider range of linguistic phenomena.

The student should either be in a computational linguistics or computer science department or have an equivalent background. Solid programming skills in Java or Python are required, previous experience with distributional semantics and with inferential statistics (t-test, ANOVA, GLM) is a plus. The work language will be English but we will work with German data.

This thesis would be carried out as a cooperation between the Department of Computational Linguistics and Phonetics in Saarbrücken and the Institut für Maschinelle Sprachverarbeitung in Stuttgart. Please email zarcone (at) coli.uni-saarland.de (if you are based in Saarbrücken) or pado (at) ims.uni-stuttgart.de (if you are based in Stuttgart) if you are interested in making an appointment and talking to us about this proposal.