As part of my PhD research I attended the RANLP 2007 conference in Borovets, Bulgaria (Recent Advances in Natural Language Processing – 2007) to present two scientific papers and a novel algorithm for measuring semantic similarity between words by searching Google and analyzing the local contexts comming for the returned page excerpts.

Improved Word Alignments Using the Web as a Corpus

With my co-authors we presented an approach for improving word alignments by combining orthographic similarity measure designed for Russian/Bulgarian with semantic similarity measure. The full paper is available here: Improved Word Alignments Using the Web as a Corpus (PDF). The presentation is also available: Improved Word Alignments – Presentation RANLP 2007 (PPT).

Cognate or False Friend? Ask the Web!

With my co-authors we presented a novel algorithm for distinguishing between cognates and false friends. The algorithm is based on improved version of our semantic similarity measure that uses the Web as a corpus. We also published an evaluation of its performance for Russian/Bulgarian. The full paper is available here: Cognate or False Friend? Ask the Web! (PDF). The presentation is also available: Cognate or False Friend? Ask the Web? (PPT). We also recorded video of my talk and the discussion: Cognate-or-False-Friend-Ask-the-Web-v1.0.avi.