Temporal recognition

[henry]

i am training perform temporal recognition for my research i am am
trying to find a way of how to use the annotated data (TimeBank) for
training. my main goal is to be able to recognize the temporal
expressions using machine learning

sample :
<EVENT eid="e33" class="OCCURRENCE" stem="own" aspect="NONE"
tense="PRESENT" polarity="POS" pos="VERB">owns</EVENT> eight
condominiums here.</s>
<s>Pak ca n't <EVENT eid="e34" class="OCCURRENCE" stem="find"
aspect="NONE" tense="PRESENT" polarity="NEG" pos="VERB">find</EVENT>
buyers.</s> <s>She <EVENT eid="e37" class="REPORTING" stem="estimate"
aspect="NONE" tense="PRESENT" polarity="POS" pos="VERB">estimates</
EVENT> her properties, <EVENT eid="e88" class="STATE" stem="worth"
aspect="NONE" tense="NONE" polarity="POS" pos="ADJECTIVE">worth</
EVENT> a hundred thirty million dollars in <TIMEX3 tid="t89"
type="DATE" value="1997-10" temporalFunction="true"
functionInDocument="NONE" anchorTimeID="t82">October</TIMEX3>, are
<EVENT eid="e91" class="STATE" stem="worth" aspect="NONE"
tense="PRESENT" polarity="POS" pos="ADJECTIVE">worth</EVENT> only half
that <TIMEX3 tid="t90" type="DATE" value="PRESENT_REF"
temporalFunction="true" functionInDocument="NONE"
anchorTimeID="t82">now</TIMEX3>.</s>

[Hector Llorens]

Hi,

In my opinion you should start reading the papers of the participating systems in TempEval-2.
Perhaps, in your early beginning it could be easier to use the TempEval-2 data wich is already tokenized (instead of TimeBank).

If your are planning to use NLP tools for PoS tagging, lemmatizing and syntactic parsing you will find the tokenized output in a tab format (for example, following TempEval-2 data) very similar to the feature vectors most machine learning techniques require as input and then useful for an easy start.

Hope this helps.
Hector

2011/1/27 henry <henryk...@gmail.com>