Utrecht University invites excellent candidates to apply for the postdoc position in computational linguistics. The application deadline is the 8th of September 2019.
Successful applicants will be participating in the project ‘A composition calculus for vector-based semantic modelling with a localization for Dutch’. Focusing on empirical data from Dutch, the project aims to combine a vector-based modelling of lexical semantics with a typelogical compositional account of how word meanings are combined to obtain interpretations for larger linguistic units. The project is funded by the NWO Open Competition programme and will be conducted at the Utrecht Institute of Linguistics OTS, at Utrecht University, the Netherlands. Principal investigator (PI) is Michael Moortgat.
The project consists of the following three subprojects. P1 and P2 are currently in progress. The present job vacancy concerns P3.
P1. *Semantic structures for enriched compositional distributional models*. This project works out the logical and mathematical foundations of the proposed grammar architecture and its compositional vector-based interpretation. A first aim is to show how the derivations of a state-of-the art typelogical syntax model can be mapped to an interpretation in finite dimensional vector spaces and linear maps. Secondly, we investigate enriched distributional models that exploit the properties of the syntactic source logic more fully.
P2. *Typelogical grammar induction for vector-based semantics from Dutch treebank data*. Whereas P1 develops a language-independent framework for compositional distributional semantics, the present subproject develops a localization of the abstract model for Dutch. The project applies machine learning techniques to obtain the required lexical entries and the associated vectorial representations from annotated corpora. For the distributional interpretation, it employs a linguistically informed notion of `context’, based on the dependency relations of words.
P3. *Empirical validation of the extended vector-based semantic model*. This project evaluates the abstract grammatical model on Dutch empirical data, and conducts a number of experiments in order to measure the performance of the model as compared to alternative approaches. The focus is on the structural particularities of the Dutch language that pose a challenge to a compositional vector-based interpretation.
For a more complete project description, details on terms of employment, salary, requirements and application format, please visit the Utrecht University job advertisement link for 21/08/2019.
For additional information please contact Michael Moortgat