... more fine-grained event duration in-
formation, viz., the most likely temporal units of
event durations (cf. (Rieger 1974)’s
ORDER-
HOURS
, ORDERDAYS).
For each original event annotation, ...
learning event durations.
3.1 Local Context
For a given event, the local context features in-
clude a window of n tokens to its left and n to-
kens to its right, as well as the event i...
... M(h)): remove h from L
6: else if (M(h) M(h
∗
)): remove h from L; set
h
∗
= h
7: end for
8: Add h
∗
to Frontier Set F
9: for each h in L do
10: if (M(h
∗
) M(h)): remove h from L
11: end ... here; testset trends are similar but not
included due to space constraints.
7
5
from www.kecl.ntt.co.jp/icl/lirg/ribes
6
from www.umd.edu/
˜
snover/tercom
7
An aside: For comparing optimiza...
... illustrate
how our feature sets are extracted from CONLL-
formatted data (Nivre et al., 2007). The CONLL
format is a common language for comparing output
from dependency parsers: each lexical item ... from the VALEX lexicon’s data set,
which was used in previous studies (Sun and Korho-
nen, 2009; Preiss et al., 2007) and facilitates com-
parison with that resource. This data set is draw...
... through language links
for training and testing. We obtained a total of
155,310 article pairs, from which we then
randomly selected 13,150 and 2,181 titles as seeds
to obtain the training and ... data. Since we are
using Wikipedia bilingual titles as the gold
standard, we exclude any snippets from the
wikipedia.org domain, so that we are not using
Wikipedia article content in both...
... 2012.
c
2012 Association for Computational Linguistics
Learning the Latent Semantics of a Concept from its Definition
Weiwei Guo
Department of Computer Science,
Columbia University,
New York, NY, ... discover more overlapping words. How-
ever, exact word matching is lossy. Below are two
definitions from WN:
bank#n#1: a financial institution that accepts deposits
and channels the money into...
... follows closely
from their work. Their approach can incorpo-
rate expert supervision into the reward function
in a similar manner to this paper, but is also able
to learn effectively from environment ... direc-
tions. Where traditional models learn
from linguistic annotation or word distri-
butions, our approach is grounded in the
world, learning by apprenticeship from
routes through...
... taxo-
nomic relation (term pairs are taken from Word-
Net (Miller et al., 1990)); then they parse the sen-
tences, and automatically extract patterns from the
parse trees. Finally, they train ... of
word lattices that we use to model tex-
tual definitions. Lattices are learned from
a dataset of definitions from Wikipedia.
Our method is applied to the task of def-
inition and hypernym ex...
... dataset
of Pang and Lee (2004), which contains subjective
sentences from movie review summaries and objec-
tive sentences from movie plot summaries. This task
2
Dataset and further details are ... imbalance in ratings present
in review collections. This weighting prevents the
overall distribution of document ratings from affect-
ing the estimate of document ratings in which a par-
ticul...
... phrase.
3.2 Grammar Extraction
From every word-aligned sentence-pair and its la-
bel chart, we extract SCFG rules as those of Figure
2. Binary rules are extracted from adjoining syn-
chronous spans ... and Chinese as tar-
get. The data for the first three language pairs are
derived from parliament proceedings sourced from
the Europarl corpus (Koehn, 2005), with WMT-
07 development and...
... the OOV detector
training set is different from the LVCSR training set.
We also use a hybrid LVCSR system, combin-
ing word and sub-word units obtained from ei-
ther our approach or a state-of-the-art ... lexicon includes
resulting sub-words – ranging from unigrams to 5-
gram phones, and the 83K word lexicon.
5.3 Evaluation
We obtain confusion networks from both the word
and hybrid LV...