NLP for CP
Addressing Constraint Programming with Natural Language Processing
Home
Resources
Publications
Correct
predictions are in
blue
. If we detect only a subset of a labelled sentence, we highlight the caught part as
blue
, the missing part
light blue.
False positives
are in
green
and
false negatives
are in
red
.
Problem 115 (Tail Assignment) — Constraint detection
Tail
Assignment
is
the
problem
of
deciding
which
individual
aircraft
-LRB-
identified
by
its
tail
number
-RRB-
should
cover
which
flight
.
Each
aircraft
is
thus
assigned
a
route
consisting
of
a
sequence
of
flights
,
and
possibly
other
activities
such
as
maintenance
,
to
perform
.
Tail
Assignment
deals
with
individual
constraints
,
flights
which
are
fixed
in
time
,
as
well
as
individual
rules
for
each
tail
.
The
planning
period
is
typically
one
month
.
The
purpose
is
to
really
create
a
solution
that
is
possible
to
operate
,
satisfying
all
rules
and
regulations
.
The
most
basic
rules
are
rules
which
only
depend
on
two
flights
,
so-called
connection-based
rules
.
For
example
,
there
must
be
a
certain
minimum
buffer
time
between
a
landing
and
the
next
take-off
.
Another
important
set
of
constraints
are
the
flight
restriction
rules
,
which
forbid
certain
aircraft
to
operate
certain
flights
.
There
can
be
many
reasons
for
the
restriction
-
there
can
be
a
curfew
for
the
arrival
airport
and
some
aircraft
,
because
the
aircraft
violates
noise
or
environmental
restrictions
.
But
there
can
also
be
more
down-to-earth
reasons
,
like
the
aircraft
not
having
the
required
in-flight
entertainment
system
or
extra
fuel
tanks
required
for
a
long
flight
.
Either
way
,
the
result
is
that
an
aircraft
is
restricted
from
operating
a
flight
.
Finally
,
there
are
the
maintenance
rules
.
Aviation
authorities
require
that
all
aircraft
undergo
various
types
of
maintenance
activities
regularly
.
There
are
many
maintenance
types
,
depending
om
aircraft
type
,
registration
country
,
and
airline
.
Typically
,
the
rules
specify
that
aircraft
must
undergo
maintenance
every
X
hours
,
or
every
Y
landings
.
Airlines
often
also
require
that
their
aircraft
return
to
a
maintenance
base
frequently
,
even
if
no
maintenance
is
done
,
to
increase
robustness
in
case
disruptions
occur
.
These
rules
typically
specify
that
aircraft
must
come
back
to
a
maintenance
base
every
Z
days
.
The
normal
representation
of
the
Tail
Assignment
problem
is
in
terms
of
a
flight
network
.
In
the
flight
network
,
each
node
represents
a
flight
,
or
some
other
activity
such
as
a
preassigned
maintenance
activity
for
specific
aircraft
,
and
each
arc
represents
a
connection
between
two
flights
or
activities
.
For
example
,
if
operating
flight
f
followed
by
flight
f
'
is
allowed
according
to
connection
rules
,
the
connection
from
f
to
f
'
is
considered
legal
,
and
the
flight
network
will
contain
an
arc
between
nodes
f
and
f
'
.
Since
we
are
solving
a
dated
problem
,
where
flights
are
fixed
in
time
,
there
are
carry-in
activities
in
the
beginning
of
the
period
representing
the
last
flights
operated
by
each
aircraft
in
the
previous
planning
period
,
and
the
network
is
acyclic
.
The
goal
is
now
to
find
paths
-LRB-
routes
-RRB-
through
the
network
for
all
aircraft
,
starting
at
the
carry-in
activities
,
such
that
all
flight
nodes
are
covered
exactly
once
,
and
all
rules
are
satisfied
.
Problem 115 (Tail Assignment) — Detection of the decisions and objects to be modeled
Tail
Assignment
is
the
problem
of
deciding
which
individual
aircraft
-LRB-
identified
by
its
tail
number
-RRB-
should
cover
which
flight
.
Each
aircraft
is
thus
assigned
a
route
consisting
of
a
sequence
of
flights
,
and
possibly
other
activities
such
as
maintenance
,
to
perform
.
Tail
Assignment
deals
with
individual
constraints
,
flights
which
are
fixed
in
time
,
as
well
as
individual
rules
for
each
tail
.
The
planning
period
is
typically
one
month
.
The
purpose
is
to
really
create
a
solution
that
is
possible
to
operate
,
satisfying
all
rules
and
regulations
.
The
most
basic
rules
are
rules
which
only
depend
on
two
flights
,
so-called
connection-based
rules
.
For
example
,
there
must
be
a
certain
minimum
buffer
time
between
a
landing
and
the
next
take-off
.
Another
important
set
of
constraints
are
the
flight
restriction
rules
,
which
forbid
certain
aircraft
to
operate
certain
flights
.
There
can
be
many
reasons
for
the
restriction
-
there
can
be
a
curfew
for
the
arrival
airport
and
some
aircraft
,
because
the
aircraft
violates
noise
or
environmental
restrictions
.
But
there
can
also
be
more
down-to-earth
reasons
,
like
the
aircraft
not
having
the
required
in-flight
entertainment
system
or
extra
fuel
tanks
required
for
a
long
flight
.
Either
way
,
the
result
is
that
an
aircraft
is
restricted
from
operating
a
flight
.
Finally
,
there
are
the
maintenance
rules
.
Aviation
authorities
require
that
all
aircraft
undergo
various
types
of
maintenance
activities
regularly
.
There
are
many
maintenance
types
,
depending
om
aircraft
type
,
registration
country
,
and
airline
.
Typically
,
the
rules
specify
that
aircraft
must
undergo
maintenance
every
X
hours
,
or
every
Y
landings
.
Airlines
often
also
require
that
their
aircraft
return
to
a
maintenance
base
frequently
,
even
if
no
maintenance
is
done
,
to
increase
robustness
in
case
disruptions
occur
.
These
rules
typically
specify
that
aircraft
must
come
back
to
a
maintenance
base
every
Z
days
.
The
normal
representation
of
the
Tail
Assignment
problem
is
in
terms
of
a
flight
network
.
In
the
flight
network
,
each
node
represents
a
flight
,
or
some
other
activity
such
as
a
preassigned
maintenance
activity
for
specific
aircraft
,
and
each
arc
represents
a
connection
between
two
flights
or
activities
.
For
example
,
if
operating
flight
f
followed
by
flight
f
'
is
allowed
according
to
connection
rules
,
the
connection
from
f
to
f
'
is
considered
legal
,
and
the
flight
network
will
contain
an
arc
between
nodes
f
and
f
'
.
Since
we
are
solving
a
dated
problem
,
where
flights
are
fixed
in
time
,
there
are
carry-in
activities
in
the
beginning
of
the
period
representing
the
last
flights
operated
by
each
aircraft
in
the
previous
planning
period
,
and
the
network
is
acyclic
.
The
goal
is
now
to
find
paths
-LRB-
routes
-RRB-
through
the
network
for
all
aircraft
,
starting
at
the
carry-in
activities
,
such
that
all
flight
nodes
are
covered
exactly
once
,
and
all
rules
are
satisfied
.
Back to list