Saturday, January 15, 2011

Do's and Don'ts: Organization of Research Talk

I have been asked this question (i.e., how to organize research talk through slides) by people from sometime and I have been replying them in private communication (i.e., over chat, emails and other means). However, now I have decided to write a blog post on this subject for greater audience.

Actually, there is no perfect answer to it but there are few tips which may be meaningful and applicable after some adjustment per area of research (it also depends upon maturity of research which is being presented).

Some tips regarding a research talk (presented by slides) with do's and don'ts are as follows.

  1. Never paste a flat abstract in your slides ever, because no body is ever entitled to read it like a paper during your talk.
  2. Never copy sentences, instead try to phrase it as concepts because slide is not a collection of flat sentences. It is expected to be organization of concepts, nobody would read it like a book in your talk (neither they have time nor it is expected from a talk).
  3. Never put other's content without a reference. It would be better if you put references by hinting year of publication such as BP98 or YZ10 so that viewer can understand how latest is that work. References [1], [2] etc is not helpful for talk. Remember talk is different from book or paper.
  4. Never fill too much space with content in a single slide, thats not understandable during talk (no body reads that much in one slide)
  5. Never discuss a concept to such depths that it takes more than a minute (approx.), slide is not for discussing concepts in details.
  6. Never explain a concept by talk without inserting it into your slide. People follow your talk by looking to your slides.
  7. Never use bullet or numbered list if there is no other parallel concept to that. Definition of apple is neither a bullet nor a numbered list.

  1. Always put slide numbers so that people could refer to page number later for questioning.
  2. Always put contents in terms of conceptual hierarchy.
  3. Always put two parallel concepts at same level of bullets. Apple is parallel bullet to orange if focus is fruits.
  4. Always put sub-concepts as sub-bullets as per their logical associativity. Apple may have sub-bullets such as green apple and red apple.
  5. Always use numbered lists if concepts are ordered otherwise use bullets. Algorithm is sequential flow so use numbered list.
  6. Always remember every conceptual abstraction should be completely meaningful such as overview and observation should be 100% mappable to concepts discussed inside.
  7. Always use "(1/x)" style (or similar) for concepts that expands on more than one slide, where x represents maximum number of pages till which the concept is expanded.
  8. Always put supplementary notes at the end of slides, such that if a question comes little off from slides then you can take help from supplementary slides (if needed).
  9. Always use consistent conventions such as Fig. or Figure. Consistent means, Figure and Fig. both cannot be use in complete slide.
  10. Always read your slides after preparing it completely. Take a print out and carefully examine it, you will generally find some errors in it.
  11. Always, begin your slides, by exposing the contents of your talk. Main contents should be very general: no unknown term that requires explanation could be your main heading of your contents except title of your thesis. Following is one good example.
  1. Introduction: It covers background work i.e., basic definition of objects which serves valid for your work (preliminary definitions) and then make connections among discussed objects in a story line as evolution of your focused area. Finally, link your problem statement (as natural evolution to the subject). Motivation and Goal is extremely important in this part followed by contribution of your work (which is to be explained in coming parts of your talk).
  2. Related Work: In this part, you are required to present previous related work in your area of research. You are expected to defend your idea against presented related works in later part (i.e., experimental part) in order to compare your results (i.e., effectiveness of your work). If not all then some related works are extremely important for comparing from your work.
  3. Your Work or Title borrowed from Thesis: In this part, you are expected to explain your idea and link all the concepts together as if they were pieces of puzzle fitting together. This is your part of talk i.e., fresh and new from other works.
  4. Evaluation of your work/Experimental Evidence: In this part, you are required to show the comparison of your work with other established works in order to justify your contribution.
  5. Conclusion: In last, you are supposed to summarize your findings.
Following is the link of my defense of Masters by Research (which took place on 16 Dec. 2010). You can take some help from it as sample template (i.e., how a talk could be organized)

I also wrote a post as a small talk on "writing" which may be useful for understanding the common problems while transferring ideas through writing.

Maintaining above practice for preparing slides would provide you inside worth of your own research work i.e., you can judge where you stand in terms of actual worth of your own work (approx.), therefore by preparing good slides one can always help himself or herself before anyone else. As a researcher, you are expected to be clear in transferring your ideas by speech. Making good slides is a science which must be mastered by a researcher.

If you find this post interesting please share it with your friends. By sharing it with people it will contribute either as feedbacks (correcting, adding, questioning etc.,) for me or some knowledge for other people. Always help the world, so that you could help most in return through the cycle.

Friday, January 7, 2011

Master's Thesis: Improving the Quality of Web Spam Filtering by Using Seed Refinement

I finally defended my Master's thesis titled "Improving the Quality of Web Spam Filtering by Using Seed Refinement"on 16th December, 2010. The thesis deals with a significant problem from the viewpoint of today's World Wide Web namely Web spam which has become a nuisance for search engines today.

The thesis proposes seed refinement techniques for four well-known web spam filtering algorithms: TrustRank, Anti-TrustRank, Spam Mass, and Link Farm Spam. The input seed is refined by maintaining an exception list in the input seed set. This proves to be helpful in decreasing false positives while increasing true positives. Additionally, in this thesis, a strategy for the succession of the modified algorithms is also proposed. These are classified into two classes: a seed refiner followed by a spam detector. Modified TrustRank (MTR) and Modified Anti-TrustRank (MATR) which are seed refiners while Modified Spam Mass (MSM) and Modified Link Farm Spam (MLFS) which are spam detectors.

Following is my Master's thesis defense presentation which I am sharing for those interested:

The full-text of the thesis can be downloaded from here or here (the journal copy). Interested students/researchers may contact me for any questions, comments and feedback. The full-text of the thesis can also be requested via email.

For Citation:
  title={Improving the Quality of Web Spam Filtering by Using Seed Refinement},
  author={Qureshi, Muhammad Atif; Yun, Tae-Seob; Lee, Jeong-Hoon; Whang, Kyu-Young},
  journal={Journal of the Institute of Electronics Engineers of Korea},