TU Berlin

Database Systems and Information Management GroupLearning semantic relations with distributional similarity

Logo FG DIMA-new  65px

Page Content

to Navigation

Short info

Candidate: Priska Herger

Advisor: Alan Akbik, Prof. Dr. Chris Biemann (TU Darmstadt)

Desired degree: Master

Abstract

Learning to detect and distinguish semantic relations is an important building block in the formalization of natural language. In my thesis I examined an approach to discriminate semantic relations between pairs of nouns by representing paradigmatic nouns with their syntagmatic contexts. A logistic regression classifier and the Chinese Whispers clustering algorithm were trained on this word space and predictions evaluated with the BLESS and SAT data. Results achieved are comparable to the state-of-the-art.

Navigation

Quick Access

Schnellnavigation zur Seite über Nummerneingabe