Presenting Your Research: Papers, Presentations, and People Marie
Presenting Your Research: Papers, Presentations, and People Marie desJardins ([email protected]) AAAI/SIGAI Doctoral Consortium February 4, 2017 Thanks to Rob Holte for permission to use some slides Who Am I? Prof. Marie desJardins A.B. Engineering, Harvard 1985
Ph.D. Computer Science, Berkeley 1992 Research Scientist, SRI, 1991-2001 Professor at UMBC, 2001-now Tenured in 2007, full professor in 2011 Associate Dean, 2015-now Married since 1985, two daughters (sophomore in college & 2nd-year med student) Like to read, sing, play piano, ski, travel, eat great food, solve crossword puzzles and logic problems My Philosophy of Success What you love What the What you do world wants well
A mentors job is to help our mentees find the intersection Advisor-Advisee Relationships Know your mentor type: Your topic vs. their topic Hands off vs. hands on Explicit expectations vs. none What about your mentor? Which type are they? What students will work well with your mentors type? Are you that type of student, or can you and your mentor adapt to each other?
AAAI-17 Workshop on Diversity in AI 4 Advisor-Advisee Relationships My style: My area/suggestions; their topics Regular (usually weekly) meetings They set goals & milestones (with my input) I expect progress & regular updates I like my students to write frequently I dont formally advise grad students until theyve taken a class or independent study Challenges: Advisors are busy! Use their time wisely! It's easier to say "OK, sounds good!" than to ask hard questions and really advise
Students get off trackwhen to intervene? (Ultimately, motivation is up to the student!) If youre being funded, does the mentor have the right to be very directive? What if the funded direction isnt helping your dissertation research? When might an advisor and advisee separate? AAAI-17 Workshop on Diversity in AI 5 Research Isnt Just Research Who cares what you do, if you never tell them? Youll need to present your ideas in various forms and venues: PEOPLE: Networking with colleagues at your institution and elsewhere PUBLICATIONS: Writing and submitting papers to workshops, conferences, and journals
PRESENTATIONS: Giving talks at workshops, conferences, and other institutions (You should also put together a website that highlights your interests and research activities) oh, and these things also provide useful experience for job interviews, not to mention valuable job skills 2/4/17 Marie desJardins -- Presenting Your Research 6 People Networking Meet people! It helps to have an objective:
Find out what research theyre currently working on Tell them what youre currently working on Find an area of common interest Learn what their visions/future directions are Suggest a new direction for research or topic for a class Whats in this interaction for you? Whats in it for them? If you know two friends, and they know two friends, and they know two friends Pretty soon you know everybody! 2/4/17 Marie desJardins -- Presenting Your Research
8 Be Prepared You need to be prepared to summarize your research For your dissertation topic, you should have a 1-minute, 5-minute, and 15-minute presentation already thought through Be an engaging speaker observe your audience to see whether you are holding their attention. Be responsive to their level and area of interest! The same goes for other projects youve been working on Be able to distinguish between your original contributions, your advisors contributions, and ideas drawn from previous research Practice with other students! 2/4/17 Marie desJardins -- Presenting Your Research 9
Publishing Writing and Submitting Papers For a doctoral dissertation, you should aim for at least a couple of good conference papers and a journal paper Lots more if you want an academic position Even more than that if you want a top-20 academic position Coauthoring helps Writing these papers is great practice for the dissertation itself (and you can reuse the material!) Where to submit? Look at publication lists of people doing research related to yours, and see where they publish Publish at the conferences that have the most interesting papers 2/4/17
Marie desJardins -- Presenting Your Research 11 Paper Writing: Strategies First, decide where you plan to submit the paper You may not finish in time, but having a deadline is always helpful Two to four months away is a good planning horizon Next, decide what you will say What are the key ideas? Have you developed them yet? What are the key results? Have you designed and run the experiments yet? Have you analyzed the data? What is the key related work? Have you read the relevant background material? Can you give a good summary of it? Now get started on the work you need to do to fill in the missing holes! Write early and often: You can (and should) write in parallel with
finishing the work! 2/4/17 Marie desJardins -- Presenting Your Research 12 Paper Writing: Design Abstract summarizes the research contributions, not the paper (i.e., it shouldnt be an outline of the paper) Introduction/motivation what youve done and why the reader should care, plus an outline of the paper Technical sections one or more sections summarizing the research ideas youve developed Experiments/results/analysis one or more sections presenting experimental results and/or supporting proofs Future work summary of where youre headed next and open questions still to be answered
Related work sometimes comes after introduction, sometimes before conclusions (depends to some extent on whether youre building on previous research, or dismissing it as irrelevant) Conclusions reminder of what youve said and why its important 2/4/17 Marie desJardins -- Presenting Your Research 13 Paper Writing: Tactics Top-down design (outline) is very helpful Bulleted lists can help you get past writers block Unless youre a really talented/experienced writer, you should use these tools before you start writing prose Neatness counts! Check spelling, grammar, consistency of fonts and notation before showing it to anyone for review
If theyre concentrating on your typos, they might miss whats interesting about the content. (More about the reviewers perspective later...) Leave time for reviews! Fellow students, collaborators, advisors, A paper is only done when its submitted... and usually not even then. 2/4/17 Marie desJardins -- Presenting Your Research 14 Authorship Who should be an author? Anyone who contributed significantly to the conceptual development or writing of the paper
Not necessarily people who provided feedback, implemented code, or ran experiments What order should the authors be listed in? If some authors contributed more of the conceptual development and/or did most/all of the writing, they should be listed first If the contribution was equal or the authors worked as a team, the authors should be listed in alphabetical order Sometimes the note The authors are listed in alphabetical order is explicitly included 2/4/17 Marie desJardins -- Presenting Your Research 15 The Review Process
Conference Reviewing Program committees Selection process Senior vs. area chair vs. regular members Paper assignments Keyword-based Self-selection All for one and one for all Decisions 2/4/17
Reaching a consensus Final decisions Conditional accepts (rare) Acceptance rates (~~~20% in good conferences/journals) Marie desJardins -- Presenting Your Research 17 Journal Reviewing Executive editor Area editor Board members or reviewers Longer decision cycle Typically higher quality, longer, and deeper reviews Decision options:
2/4/17 Accept as is Accept with minor changes Accept with major changes (subject to re-review) Reject with encouragement to resubmit Reject out of hand Marie desJardins -- Presenting Your Research 18 Where to Publish Workshops vs. conferences vs. journals
2/4/17 Length of decision cycle Quantity vs. quality Aim high! (or at least appropriately) Acceptance rate vs. time to prepare/publish Have a pipeline of ideas at different stages of development, published in successively more competitive/selective venues Marie desJardins -- Presenting Your Research 19 Purpose of a Review Evaluate the papers scientific merit Check the validity of the papers claims and evidence
Judge the papers relevance and significance Provide constructive feedback to the author 2/4/17 Marie desJardins -- Presenting Your Research 20 Typical Conference Review Form 1. How RELEVANT is this paper? 2. How SIGNIFICANT is this paper? 3. How ORIGINAL is this paper? 4. Is this paper technically SOUND? 5. How well is this paper PRESENTED? Additional comments for the author(s) 2/4/17
Marie desJardins -- Presenting Your Research 21 Knowing Your Audience: A Reviewers Perspective First, I read the title: is it in my area? (self-selection) Next, I read the abstract: is it interesting? (self-selection) Next, I skim the introduction and form my opinion about the paper Next, I read the rest of the paper looking for evidence to support my view By the time I get to Section 2, I already have a very strong opinion about whether to accept or reject. Your job is to give me the evidence I need in the title and abstract to select your paper for review, and in
the introduction to result in the right opinion! 2/4/17 Marie desJardins -- Presenting Your Research 22 Good Reviews Are... Polite Fair
Concise Clear Constructive Specific Well documented Represent the scientific community ... but you get what you get! Bad, unfair review that missed the point? Fix your paper anyway! 2/4/17 Marie desJardins -- Presenting Your Research 23 Ethical Issues
2/4/17 Multiple submissions Journal versions of conference papers Authors and author order Listing papers in your CV Marie desJardins -- Presenting Your Research 24 Rejected!! Now What? Fix the paper!
2/4/17 Read the reviews, rail and complain, berate the reviewer Calm down Read them again with an open mind Do more experiments, revise the paper, Go back to the reviews again have you addressed all the points? Have people read the revision critically Do more experiments, revise the paper, Repeat until the next deadline Marie desJardins -- Presenting Your Research
25 A Tale of Two Women Dr. X: Rejected by all but one school Failed every comp exam 9 of 10 proposals rejected No papers in top journal Negative dept. vote on tenure Dozens of papers rejected Dr. Y: PhD from Top-Ten school Passed all; best grade on one $6M of sponsored funding Assoc. Editor of top journal Tenured full prof. 100+ published papers
Failure or Success? You decide Theyre both me. AAAI-17 Workshop on Diversity in AI 26 Advice for Handling Failure Know that you're in good company Don't react to failure too quickly Learn to take useful/constructive criticism Learn not to take harmful/destructive criticism personally Learn to tell useful from harmful criticism!
Realize that other people have their own agendas/interests and aren't trying to hurt you personally Take care of yourself Know your goals, priorities, and values Lean on your friends AAAI-17 Workshop on Diversity in AI 27 Presentations Know How Long You Have How long is the talk? Are questions included? A good heuristic is 1-2 minutes per slide but it depends a LOT on the slides (anywhere from 30 seconds to 3 minutes); spend too long on one slide, and youll probably lose your audience
If you have too many slides, youll skip some orworserush desperately to finish. Avoid this temptation!! Almost by definition, you never have time to say everything about your topic, so dont worry about skipping some things! Unless youre very experienced giving talks, you should practice your timing: A couple of times on your own to get the general flow At least one dry run to work out the kinks A run-through on your own the night before the talk 2/4/17 Marie desJardins -- Presenting Your Research 29 Know Your Audience Dont waste time on basics if youre talking to an audience
in your field Even for these people, you need to be sure youre explaining each new concept clearly On the other hand, youll lose people in a general audience if you dont give the necessary background In any case, the most important thing is to emphasize what youve done and why they should care! 2/4/17 Marie desJardins -- Presenting Your Research 30 Know What You Want to Say Just giving a project summary is not interesting to most people You should give enough detail to get your interesting ideas across (and to show that youve actually solved, but not
enough to lose your audience) They want to hear what you did that was cool and why they should care Preferably, theyll hear the above two points at the beginning of the talk, over the course of the talk, and at the end of the talk If theyre intrigued, theyll ask questions or read your paper Whatever you do, dont just read your slides! 2/4/17 Marie desJardins -- Presenting Your Research 31 Preparing slides Dont just read your slides! Use the minimum amount of text necessary Use examples Use a readable, simple, yet elegant format
Use color to emphasize important points, but avoid the excessive use of color Hiding bullets like this is annoying (but sometimes effective), but Abuse of animation Dont fidget, and Dont just read your slides!
2/4/17 is a Marie desJardins -- Presenting Your Research cardinal sin! 32 How to Give a Bad Talk Advice from Dave Patterson, summarized by Mark Hill 1. Thou shalt not be neat 2. Thou shalt not waste space
3. Thou shalt not covet brevity 4. Thou shalt cover thy naked slides 5. Thou shalt not write large 6. Thou shalt not use color 7. Thou shalt not illustrate 8. Thou shalt not make eye contact 9. Thou shalt not skip slides in a long talk 10. Thou shalt not practice 2/4/17 Marie desJardins -- Presenting Your Research 33 Some Useful Resources Any Questions? Some useful resources:
Writing: Lynn DuPre, Bugs in Writing Strunk & White, Elements of Style Giving talks: Mark Hill, Oral presentation advice Patrick Winston, Some lecturing heuristics Simon L. Peyton Jones et al., How to give a good research talk Dave Patterson, How to have a bad career in research/academia (An earlier, longer version of) these slides: http://www.cs.umbc.edu/~mariedj/talks/presenting-research-dc-jul05.ppt 2/4/17 Marie desJardins -- Presenting Your Research 34
"Willow and Ginkgo" by Eve Merriam. The willow is like an etching,Fine-lined against the sky.Then ginkgo is like a crude sketch,Hardly worthy to be signed.The willow's music is like a soprano,Delicate and thin.The ginkgo's tune is like a chorusWith everyone...
Tvrtka Teradata predstavlja sustav upravljanja bazama podataka stvoren za pomoć u odlučivanju 1988. IBM-ovi znanstvenici Barry Devlin i Paul Murphy razvijaju tkz. poslovno skladište podataka 1991. Bill Inmon objavljuje knjigu: Building the Data Warehouse 1991.
Dependency Load. The proportion of a country's population that is typically not in the labour force. This is used to show how much strain there is on the working class (15-64) to get jobs and support these people. Includes. Under...
Frequentis offers initial further migration services for customers who want to connect to a major AIM data repository like EAD. Frequentis plans that together with the customer and co-ordinates all technical pre-requisites that the migration process is successful
FM employees found the semi-trailer across the road at another truck stop. FM employee asked if the customer needed medical treatment (ambulance) which the customer declined. ... Want to Add 500 points to your card. Making Safety A Family Affair.
Ready to download the document? Go ahead and hit continue!