More Related Content Similar to From Work To Word (20) More from Walid Maalej (17) From Work To Word1. From
Work
to
Word:
How
Do
So/ware
Developers
Describe
Their
Work?
Walid
Maalej,
Technische
Universität
München
Hans‐Jörg
Happel,
FZI
Research
Center
Karlsruhe
2. Outline
1
MoMvaMon
2
Research
SeKng
3
Research Results
4
Conclusion
and
Next
Steps
©
W.
Maalej,
Mai
09
From
Work
To
Word
2
3. WHY
Do
Developers
Describe
Their
Work?
RaMonale
Capture
1
ReflecMon
&
Experience
Capture
• Remember
status
when
conGnuing
postponed
work
• Log
decisions
and
why
they
have
been
taken
• Capture
experiences
on
problems
and
their
soluGons
in
this
parGcular
way
• Reason
about
previous
2
decisions
by
reading
Purposes
of
Work
Status
Awareness
the
work
descripGons
DescripMon
• Help
collaborators
in
4
distributed
projects
to
know
about
the
status
of
work
Controlling
• Provide
a
short,
human
• Report
work
done
in
a
period
of
Gme
readable
summary
of
• Control
effort
and
cost
(e.g.
for
adding
one
feature)
3
changes
©
W.
Maalej,
Mai
09
From
Work
To
Word
3
4. WHERE
Do
Developers
Describe
Their
Work?
Personal
note
Commit
message
ArMfacts
Issue
report
comments
including
work
descripMons
Social
media
Time
sheet
But:
How do
developers
describe
work?
This
is
the
goal
of
our
exploraMve
study
©
W.
Maalej,
Mai
09
From
Work
To
Word
4
5. Outline
1
MoMvaMon
2
Research
SeKng
3
Research Results
4
Conclusion
and
Next
Steps
©
W.
Maalej,
Mai
09
From
Work
To
Word
5
6. Our
Research
QuesMons
On
Work
DescripMon
Content Analysis Time Metadata
Vocabulary
usage,
Session
duraMon
and
similariGes
and
terms
descripGon
frequency
frequency
Logging
day
Mme
and
Work
categories
and
relaGon
to
descripGon
proporGon
of
acGviGes
quality
DescripGon
pa_erns
and
Pseudo
descripGons,
common
templates
only
Gme
metadata
Referenced
arGfacts
Required
effort
©
W.
Maalej,
Mai
09
From
Work
To
Word
6
7. Data
Sets
Collected
in
Different
Contexts
Data
set
Summary
Number
of
Number
of
Represented
developers
descripMons
period
Developers’
Gme
8
years
MyComp
cards
in
a
German
25
38,045
(2001
–
2009)
soWware
company
Commit
messages
of
15
years
Apache
1,949
747,403
all
Apache
projects
(1994
–
2009)
Subjects‘
personal
notes
in
a
field
study
10
days
Eureka
21
115
with
5
European
(2008)
companies
©
W.
Maalej,
Mai
09
From
Work
To
Word
7
8. Outline
1
MoMvaMon
2
3
Research Results
4
Conclusion
and
Next
Steps
©
W.
Maalej,
Mai
09
From
Work
To
Word
8
9. Vocabulary
Usage
and
Term
Frequency
Rank
Mycomp German Apache Eureka English
Term
Not 2 17 4 1 29
For/ Since 7 184 23 212 906
Change 3 911 26 214 333
Review 1 >10,000 13 - 2,275
Problem 24 461 10 9 239
Now 12 576 14 58 808
Done 17 4,454 300 59 606
Work
descripMons
extensively
include:
• NegaMve
formulaMons
• JusMficaMons
and
argumentaMons
• Problems
that
caused
the
work
©
W.
Maalej,
Mai
09
From
Work
To
Word
9
10. Described
Categories
of
Work
• Not
only
acMviMes
“with
work
products”
are
described
• Granularity
is
more
“ediMng‐”
rather
than
acMvity‐oriented
©
W.
Maalej,
Mai
09
From
Work
To
Word
10
11. Work
DescripMon
Pa_erns
1.
ArgumentaMon
Pa_erns
Pa#ern ::= <Ac-on> 'concerning' | 'performed on' <Ar-fact>
'for'|'since'|'because' <Cause descrip-on>|<Reference to cause>
2.
Status
Pa_erns
Pa#ern ::= <Ar-fact> 'works now'
Pa#ern ::= 'Problem'|'Bug' <Reference to issue>|<Problem descrip-on>
'fixed'|'solved' <Solu-on> ['reviewed with' <Colleague>]
3.
Experience
Sharing
Pa_erns
Pa#ern ::= 'If' <Context> 'then' <Experience>
©
W.
Maalej,
Mai
09
From
Work
To
Word
11
12. CreaMon
Time
and
Session
DuraMon
• Delayed
descripMons
are
shorter
than
immediate
ones
• The
mean
of
session
duraMons
is
between
30
and
90
min.
• Developers
entered
between
8
and
12
descripMons
per
work
day
©
W.
Maalej,
Mai
09
From
Work
To
Word
12
13. Work
DescripMon
Quality
and
Effort
10%
of
pseudo
descripMons
• Developers
don’t
have
Mme
or
moMvaMon
to
describe
10%
of
the
sessions
• 3
–
6%
of
developer
Mme
is
spent
for
describing
work
(30
min.
/
day
)
©
W.
Maalej,
Mai
09
From
Work
To
Word
13
14. Outline
1
MoMvaMon
2
Research
SeKng
3
Research Results
4
Conclusion
and
Next
Steps
©
W.
Maalej,
Mai
09
From
Work
To
Word
14
15. Summary
of
the
Talk
1
We
conducted
an
exploratory
study
on
how
developers
describe
their
work
using
real
world
data
2
A
considerable
amount
of
effort
is
spent
to
describe
work,
with
oWen
empty
or
pseudo‐descripGons:
automaGon
would
pay
off
3
We
found
similariGes
in
descripGon
contents:
part
of
the
descripGon
can
be
automated
by
observing
interacMons
4
We
found
similariGes
in
the
Mme
metadata:
the
work
day
can
be
sessionized
automaGcally
©
W.
Maalej,
Mai
09
PotenMals
and
Challenges
of
RS
in
SD
15
18. You
are
welcome
to
join!
Contact:
Walid
Maalej
Hans‐Jörg
Happel
TUM
FZI
maalejw@cs.tum.edu
happel@fzi.de
©
W.
Maalej,
Mai
09
From
Work
To
Word
18