Information Technology Reference
In-Depth Information
Rule set 1
:
Section-event Rules
In patient records, we need to determine the TLink types between the events
and identify the admission/discharge times. And we designed seven section-event
rules (R.SE) to determine the TLink type of the section-event pair.
R.SE.1:
inHospitalCourseSection
(
s
i
)
∧
inSentence
(
e
j
,s
i
)
TLink
(
e
j
,t
discharge
,
“
BEFORE
)
R.SE.2:
inClinicalHistorySection
(
s
i
)
TLink
(
e
j
,t
admission
,
“
AFTER
)
⃒
∧
∧
inSentence
(
e
j
,s
i
)
TLink
(
e
j
,t
admission
,
“
OV ERLAP
)
TLink
(
e
j
,t
discharge
,
“
BEFORE
)
⃒
∧
R.SE.3:
inAdmissionSection
(
s
i
)
∧
inSentence
(
e
j
,s
i
)
TLink
(
e
j
,t
admission
,
“
OV ERLAP
)
TLink
(
e
j
,t
discharge
,
“
BEFORE
)
⃒
∧
R.SE.4:
inDischargeSection
(
s
i
)
∧
inSentence
(
e
j
,s
i
)
TLink
(
e
j
,t
admission
,
“
AFTER
)
TLink
(
e
j
,t
discharge
,
“
OV ERLAP
)
⃒
∧
R.SE.5:
inOtherSection
(
s
i
)
∧
inSentence
(
e
j
,s
i
)
TLink
(
e
j
,t
discharge
,
“
OV ERLAP
)
R.SE.6:
inClinicalHistorySection
(
s
i
)
TLink
(
e
j
,t
admission
,
“
AFTER
)
⃒
∧
∧
hasAdmissionKeyword
(
s
i
)
TLink
(
e
j
,t
admission
,
“
BEFORE
)
TLink
(
e
j
,t
discharge
,
“
BEFORE
)
⃒
∧
R.SE.7:
inClinicalHistorySection
(
s
i
)
∧
hasPastWord
(
s
i
)
TLink
(
e
j
,t
discharge
,
“
BEFORE
)
where
s
i
is the
i
th sentence,
e
j
is the
j
th event,
t
admission
is the admis-
sion time,
inSentence
(
e
j
,s
i
)meansthe
j
th event is in the
i
th sentence, and
inHospitalCourseSection
(
s
i
)meansthe
s
i
sentence is in hospital course sec-
tion. The adminision keywords are ”admission”,”admit” and ”present”. For
instance, R.SE.1 means
s
i
is in hospital course section in the patient record
and there is an event
e
i
in
s
i
.
e
i
will be
TLink
(
e
j
,t
admission
,
“
AFTER
)and
TLink
(
e
j
,t
discharge
,
“
BEFORE
).
Rule set 2
:
Within-sentence Rules
Events and times appearing in the same sentence are very likely to have the
same TLink type. We developed five rules (R.WS) to classify their TLink types.
R.WS.1:
isEventType
(
e
i
,
“
PROBLEM
)
∧
isEventType
(
e
j
,
“
TREATMENT
)
⃒ TLink
(
e
i
,e
j
,
“
AFTER
)
R.WS.2:
isEventType
(
e
i
,
“
CLINICAL
D
EPT
)
TLink
(
e
j
,t
admission
,
“
BEFORE
)
⃒
∧
∧
isEventType
(
e
j
,
“
TREATMENT
)
TLink
(
e
i
,e
j
,
“
OV ERLAP
)
R.WS.3:
isEventType
(
e
i
,
“
PROBLEM
)
⃒
∧
isEventType
(
e
j
,
“
TEST
)
TLink
(
e
i
,e
j
,
“
BEFORE
)
R.WS.4:
isEventType
(
e
i
,etype
)
⃒
∧
TLink
(
e
i
,e
j
,
“
OV ERLAP
)
R.WS.5:
inSentence
(
e,s
i
)
isEventType
(
e
j
,etype
)
⃒
∧
inSentence
(
t, s
i
)
∧
TLink
(
e,t,
“
OV ERLAP
)
where
isEventType
(
e
i
,
“
PROBLEM
)meansthe
e
i
is the event type
“PROBLEM”, and
theNearestTime
(
e,t
) means the time
t
is the nearest time
to the event
e
.
Rule set 3
:
Cross-sentence Rules
Cross-sentence rules classify the TLink types of the event pairs which cross
sentences. According to Tang et al. [8]'s observation, there are two main types of
cross-sentence TLinks: (1) TLinks between main events in consecutive sentences,
theNearestTime
(
e,t
)
⃒
Search WWH ::
Custom Search