How to Write a Research Announcement

Paper (or How to

z Over the summer, Prof. Mitra’s and Prof.
Graduate Quickly)? Lee’s groups will have a joint DB seminar
z Goals:
z Forum for practice talks
DB Group Summer Seminar z Learn what others are working on
z Get fresh ideas from others’ works
Dongwon Lee z Find collaborators for your research
May 19, 2005 z Get to know each other

DB Seminar Talk, 2005, Dongwon Lee

What is the Goal of a

Justification Research Paper?
z I am probably a qualified person to give a talk z Disseminate your ideas to others so that
on this topic… because people appreciate/use/cite them
z I’m still STRUGGLING to publish z Graduate… Of course
z I do get rejections a lot :-( z MS: need to write thesis to graduate…
z I’m still learning from failures z Ph.D: “Publish or Perish”
z Without good publications…
z No good job, no good career
z What’s being presented here is purely my z And possibly no good life either
suggestion
z GPA: nobody cares
z Take it or leave it – upto you !! z Maintain about 3.0/4.0
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

Where to Start? Where to Start?

DB Conferences/Symposiums/Workshops (81)
ADB, ADBIS, ADBT, ADC, ARTDB, Berkeley Workshop, BNCOD, CDB, CIDR, CIKM, • Start from good ones:
CISM, CISMOD, COMAD, COODBSE, CoopIS, DAISD, DANTE, DASFAA, DaWaK,
DBPL, DBSEC, DDB, DDW, DEXA, DIWeB, DMDW, DMKD, DNIS, DOLAP, DOOD,
– DB: SIGMOD, VLDB,
DPDS, DS, EDBT, EDS, EFIS/EFDBS, ER, EWDW, FODO, FoIKS, FQAS, Future ICDE, EDBT, …
Databases, GIS, HPTS, IADT, ICDE, ICDM, ICDT, ICOD, IDA, IDC(W), IDEAL,
IDEAS, IDS, IGIS, IWDM, IW-MMDBMS, JCDKB, KDD, KR, KRDB, LID, MDA/MDM, – DB Theory: PODS, ICDT, …
MFDBS, MLDM, MSS, NLDB, OODBS, OOIS, PAKDD, PKDD, PODS, RIDE, RIDS,
RTDB, SBBD, SDM-SIAM, Semantics in Databases, SIGMOD, SSD, SSDBM, SWDB, – Data Mining: KDD, ICDM, SDM, …
TDB, TSDM, UIDIS, VDB, VLDB, WebDB, WIDM, WISE, XP, XSym
– Modeling: ER, …
DB Journals (19)
– Information Retrieval: SIGIR, CIKM, …
ACM TODS, ACM TOIS, DKE, Data Base, DMKD, DPD, IEEE Data Eng. Bulletin,
IEEE TKDE, Info. Processing and Management, Info. Processing Letters, Info. – Digital Library: JCDL, ECDL, CIKM, …
Sciences, Info. Systems, J. of Cooperative Info. Systems, J. of Database
Management, JIIS, KAIS, SIGKDD Explorations, SIGMOD Record, VLDB J. – Web: WWW, WebDB, …

The list excludes Information Retrieval and Digital Library

• Look at DBLP: http://www.informatik.uni-trier.de/~ley/db/
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

1
Where to Start? Reference Chase
z Don’t be afraid to read journal papers z Don’t trap into the “Exponential Reference
z DB field is a fast-moving discipline: Chase” problem
z Latest techniques appear in conference/workshop
z More mature work appears in journal Paper to read
z Although longer than conference version, often in queue

easier to read
z Lots of examples, figures, descriptions, …
z Examples:
z ACM TODS, ACM TOIS, VLDB J., IEEE TKDE, ACM TOIT
z ACM Computing Survey, C. ACM, SIGMOD Record, …
Papers that you read

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

How to Find the DARN

Symptoms Research Problem?
z After chasing relevant works that are z Easy but non-helpful answer:
increasing super-exponentially fast, you z Read and think and read and think and…
would feel… z Subjective but MAYBE-helpful answer
z All relevant problems are ALREADY studied z MAP approach
by someone else z MATRIX approach
Others have 1000+ history: Mathematics, Art, …
What I Call M2D2
z z DELTA approach
z Problem is too BROAD for me to tackle z DROP approach
z Divide-n-conquer

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

1. MAP Approach 2. MATRIX Approach

z To start a research, initially, you have to read a lot of
papers anyway
z While reading those, why don’t you analyze and
summarize what you’ve read and put them into your
own wording?
z Good for a survey paper – a MAP for future readers
z To be publishable, your survey must have novel
view-point, taxonomy, comprehensive analysis, or
all of them
z Good target: ACM Comp. Survey, SIGMOD Record,
ACM C.ACM, IEEE Computer, …
z Now, You have read a lot of papers
z Draw a MATRIX on a specific problem, and
map the paper that you read to cells of matrix
z At the end, non-filled cell is the missing work
that no one has done
z But wait… first make sure that:
z The hole is worthwhile to fill in
z Doable (good as my dissertation topic?)
z Value (what’s good?)

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

2
 Example: XML-Relational
Conversion Problem 3. DELTA Approach
z Arguably easiest…
Schem Cons Query View Trigg Secur Top-K Temp Spatia z Pick one paper of your interest
a traint ers ity oral l
z Read a lot – more than 10 times
XML
Æ
z Find limitations and Extend it by DELTA
Relati O O O O O O z Prove or demonstrate that
onal
(40+) (5+) zThe limitation that you pointed out is valid
Relati
zYour suggestion improved the problem by DELA
onal z The more well-known work you choose, the harder
Æ O O O O to improve, but the better for your reputation…
XML
z Eg, “E.F. Codd’s relational model is insufficient to handle
semi-structured model because…”
z The bigger the DELTA is, the better your paper
gets
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

Example: The optimal

wedding problem Possible DELTAs
z When a person has a chance to date K z Parameters fitting:
How to determine K? Estimate?
persons, the optimal wedding algorithm is: z

z How to determine C? Comparison?

z Date upto K/3 persons
z Scalability? K=10 vs. K=100,00? Sub-optimal?
z Let the best person among K/3 as B using a criteria C
z Question the assumptions:
z Start dating again from K/3+1 person, p z Monogamy vs. Polygamy vs. N-gamy? (How to find nth best
z If p is better than B using C spouse fast?)
z Stop and Marry p z Data distribution? Uniform/Poisson/Scale-free

z Otherwise, keep dating till K-th person z Application to another domain?

z System building?
z How many ways can we improve this algo?
z …
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

4. DROP Approach
Which DELTA to Choose (adopted from J. Widom’s slides)

z Pick the DELTA that is the most significant z Pick a simple but fundamental assumption
z Some criteria are: underlying traditional database systems
z Have practical values z DROP it
z Has motivational scenario as of NOW, or z Reconsider all aspects of data management
z Predicted to be useful in N years and query processing
z Many Ph.D. theses
z Non-trivial
z Prototype from scratch
z Hot topics:
z Streaming, XML, Sensor, …
From http://www-db.stanford.edu/~widom/stream.ppt

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

3
Example: Two Stanford
Projects Where to Submit?
z The LORE Project z Top-down
z Dropped assumption: z Aim at the best conference in the field
“Data has a fixed schema declared in advance” z If rejected, go to next-tier conference or symposium
z If rejected, go to next…
z Semi-
Semi-structured data (→
(→ XML)
z Bottom-up
z The STREAM Project
z Aim at workshop
z Dropped assumption: z If accepted, work more and aim at better one (symposium
“First load data, then index it, then run queries” or 2nd-tier conference)
z Continuous data streams (+ continuous queries) z After making sure that the ideas mature enough, aim at the
best conference
From http://www-db.stanford.edu/~widom/stream.ppt

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

Avoid Some Notorious Facts on Paper Reviews

Venues (adopted from J. Cho’s slides)

z “Randomly generated paper got accepted to a z 3-4 reviewers per paper

conference… MIT Prank” (slashdot, 2005)
z 10-20% acceptance rate for top-tier venues
z http://pdos.csail.mit.edu/scigen/
z Very competitive
z Eg, The World Multi-Conference on Systemics,
Cybernetics and Informatics (SCI) z Criteria
z Along your career, you will get emails from unknown z Accept/Weak Accept
venues to submit a paper, to serve as PC, etc z Neutral
z Be careful if the venue is not well-known
z Weak Reject/Reject
z Many of them are NON-REVIEWED, and Profit-Oriented
event – no academic values what so ever !! z One reject kills a paper
z At least Accept, Weak Accept and Neutral
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

How to Give a Good

About Reviewers Impression in 1-2 hours
z 15-20 papers per reviewer 1. Good introduction
z Reviewer cannot spend 5-10 hours per paper z Everyone reads it
z 20 X 10 = 200 hours = (40 hours X 5) = 5 weeks! z If not interesting, people stop reading
z No reviewers can afford this 2. Easy to read
z Give a good impression in 1-2 hours! 1. People should understand what you say
2. Easy to confuse, difficult to understand
z Impression matters the most
3. Build an excitement and a strong case
z Content comes next!
1. What is good?
4. Broad reference
WARNING: Of course, to start with, your main idea 1. Sometimes kills a paper
must be good to get into top-tier… 2. Program committee members
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

4
Good Introduction How to Write an Intro
1. What’s the problem? 1. Start with 5 bullets
2. Why is it important?
z What’s the problem?
z Mention some application, existing problems
3. Why is it difficult? z Why is it interesting?
1. Ask some not-very-obvious questions or explain naïve z …
approach
4. What others did? 2. 1-2 sentence answer to each question
5. What’s my contribution? 3. Add more content
1. Contribution bullet list (paper organization)
6. Build some excitement/surprise 4. Spend enough time on intro
1. Keep reading! You will find something interesting later 1. Bullet points enough
7. Every word should be carefully picked
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

Paper Organization (10

Easy-to-Read Paper pages)
z You can always make it complicate later 1. Introduction (2 pages)
1. Lots of examples 2. Related Work (half page)
3. Framework (2 pages)
2. Figures & Tables – Figure speaks !!
4. Main Ideas (3 pages)
z Summary of notations
5. Experiments (2 pages)
3. Define models/architecture precisely
6. Conclusion (half page)
z Explicitly write down assumptions
7. References (half page)
z Input, output, property, goal function
z Actual idea – only 3 pages!!!
4. Make a connection
z Even tiny idea can turn into a good paper if you
z Why this experiment? DEVELOP well
DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

Importance of Personal
Research Log
z Maintain personal research log
z Sketch your research ideas into a writing
z Update your ideas as time passes
z Occasionally go back to old writings
z Prepare a short review for each paper that you read
z Summary
z Pros and cons
z Limitations or problems
z If needed, contact authors and ask questions
z Usually authors are willing to discuss with their readers
Start Writing Early On…
z Even if you feel you are NOT ready yet
z Your advisor will throw away your initial draft
anyway
z Your initial submission will be rejected anyway
z But you get
z (good or bad) Experiences and learn from that
z Writing sharpens your ideas and gives more ideas
z Writing can be improved only via writing

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

5
Fabrication and Plagiarism dbworld
z “Prominent Physicist Fired for Faking Data • Be a member of dbworld newsgroup
Research: Bell Labs says scientist 'recklessly'
misrepresented work on microprocessors…” (2002, – http://www.cs.wisc.edu/dbworld/
LA Times)
z http://www.latimes.com/news/science/la-sci- – Free membership
physicist26sep26.story – Keep track of DB-related news
z “Constantinos V. Papadopoulos got caught
plagiarism at EUROPAR (1995)… 7 papers
published and 8 under submission… all plagiarized
from Technical Reports…”
z http://www.sics.se/europar95/plagiarism.html

z NEVER, EVER, do these – professional suicide !!

DB Seminar Talk, 2005, Dongwon Lee DB Seminar Talk, 2005, Dongwon Lee

References (available at)

http://nike.psu.edu/resources/advice/
z [2002] How to write a paper?, Junghoo Cho, UCLA
z [1996] David Dill's Advice on Choosing an Advisor (or) How to
Survive as a Grad Student, David Dill
z [1996] How to Survive as a Graduate Student, Brian Noble,
David Dill, Benli Pierce, Jay Sipelstein, Jonathan Shewchuck
z [1997] How to Choose a Thesis Advisor, Michael C. Loui
z [????] How to have your abstract rejected, Mary-Claire van
Leunen and Richard Lipton
z [1994] Dissertation Advice, Olin Shivers
z [1999] Advice for Finishing that Damn Ph.D., Daniel M. Berry
z [1999] So long, and thanks for the Ph.D.!, Ronald T. Azuma
z [2001] How to Have a Bad Research Career, David A. Patterson

DB Seminar Talk, 2005, Dongwon Lee