TY  - RPRT
ID  - leich2008studienarbeit
T1  - Comparison of Word Sense Disambiguation Methods
A1  - Leich, Marcus
Y1  - 2008
T2  - Otto-von-Guericke-Universität Magdeburg
UR  - http://fuzzy.cs.uni-magdeburg.de/aigaion/attachments/leich2008studienarbeit.pdf
N2  - This thesis deals with the comparison of different disambiguation techniques that do
not require any form of preliminary learning. he maximum relatedness algorithm, which fulfils this criteria, is implemented together with a set of relatedness measures. By running the algorithm with these different measures, one at a time, a set of test samples is disambiguated. Results achieved, using these relatedness measures, are to be improved by appliance of the one-sense-per-discourse heuristic. Furthermore, disambiguation results of different measures are merged to combine different desired characteristic that single
measures exhibit. We propose different combination techniques and compare performance to the uncombined results.
ER  -