This webpage contains samples of our research efforts. For more detailed information and scientific publications, please visit the pages of the individual researchers.

Automated disambiguation

Existing programs for automated translation are often far from perfect. One tough problem is lexical disambiguation: how is a system supposed to know that "Pearl Jam" should not be translated and how it is supposed to know whether "rock" should be interpreted as a stone or a kind of music? To help solve this problem, we are developing a fully automated word sense disambiguation system.

Computational linguistics for African languages

The newest dictionaries are based on large collections of texts. At the AUGent we are working on learning techniques to automatically recorgnize Northern Sotho word classes. For example "kae" can be an adverb as well as an adjective. With the new software the developers of the South African Northern Sotho - English dictionary can automatically search for sample sentences with "kae" as an adverb.

Research Paper Recommendation

Given the proliferation of published research results, recommending scientific papers to researchers may provide a useful complement to traditional literature search. Various approaches may be taken to automate this task, including collaborative filtering (e.g. based on CiteULike.org or Bibsonomy.org), citation analysis (e.g. PageRank, HITS, etc.) and content-based (CB) approaches. We mainly focus on the CB paradigm, and study methods that can measure the similarity between scientific papers, or between a paper and a researcher. In particular, we investigate how baseline techniques for estimating similarity can be improved by taking various forms of meta-data about the scientific papers into account.