AnttiVirolainen T106.5820SeminaronDistributedSystems - PowerPoint PPT Presentation

Antti
Virolainen
 T‐106.5820
Seminar
on
Distributed
Systems



Describing
everyday
services


Context‐aware
content
classification


Context‐aware
content
classification



How?
 
 


Why?
 Where?
 
 


 



How?
 
 


Why?
 Where?
  Otaniemi
campus
 
 


 
  Changing
everyday
services
 The
Kassi
project


Why?
  Service
requestor
  Service
provider
  Critical
success
factor
for
Kassi
is
that





 They
Find
Each
Other
  Requires
easy
and
usable









 description
system
for
services


How?
  Describe
in
three
dimensions
  Get
clever
suggestions
based
on
context


 Make
the
big
divisions
  Are
you
offering
or
needing
something?
  Is
it
a
service
or
a
physical
object?
  Are
you
motivated
by
 reputation,
virtual
 currency
or
cold
cash?
  There
is
always
an
answer
to
all
the
questions


We
do
not
know
all
use‐cases
  A
fixed
category
set
is
not
future‐proof
  How
can
the
users
classify
the
services
that
 we
can’t
think
now?
  A
service
can
contain
anything
  Like
a
photograph
in
Flickr
  A
 folksonomy
 lets
the
users
decide
the
vocabulary


 Descriptive
labels
  Free
vocabulary
  Labeling
assisted
 based
on
context
 history 


 Who
do
you
want
to
read
your
ad?
  Friends?
  A
specific
group?
  Neighbors?
  Anyone
near
the
local
supermarket?
  What
ads
do
you
want
to
read?
  Filters
work
for
the
searcher/browser
too


The
context
information
  Data
from
the
phone
  Compare
with
earlier
recordings
  Suggest
similar


ZoneTag’s
logic
 Same
context 
 means 
 higher
probability
for
same
tags


Matching
the
context
with
history
  Location
  Social
distance
  Time
 • Of
the
day
 • Of
the
year


Example
case
  A
library
at
the
campus


Example
case
  Using
Kassi
with
mobile
phone


Example
case
  Finding
the
book


 Three‐dimensional
describing
of
everyday
 services
  Categories
  Keywords
(tags)
  Filters
(restrictions)
  Assisted
by
context‐aware
suggestions