Abstract:
In this paper, we present the preliminary experiments
of unsupervised dependency parsing on raw
segmented and part-of-speech (POS) tagged corpus
of Myanmar language. This experiment is aimed to
support building treebank and get ann-otated corpus
with dependency structures of My-anmar words. The
referenced word dependency schemes are also
explained. We present the expr-imental results on
trees of unsupervised parsed annotated corpus in
terms of unlabeled and labe-led attachment scores
(UAS and LAS) by UDPi-pe 89.79 % and 85.56% for
test and 98.25% and 97.89% for trained data
respectively.
