This paper (Mining Sentiment Classification from Political Web Logs) contains a list of political blogs.
Treat perfection like a process, not an achievable state. Perfectionism is crippling to productivity. I’ve known academics that can’t even start projects because of perfectionism. I know some academics that defend their lack of productivity by proudly proclaiming themselves to be perfectionists. I’m not so sure one should be proud of perfectionism. I don’t think it’s bad to want perfection; I just think it’s unrealistic to expect it.
The metric academics need to hit is “good enough,” and after that, “better than good enough,” if time permits. Forget that the word perfect exists. Otherwise, one can sink endless amounts of time into a project long after the scientific mission was accomplished. One good-enough paper that got submitted is worth an infinite number of perfect papers that don’t exist.
The publication structure of computer science even rewards the iterative process, as I’m sure it does in other fields as well.
- Mold an idea until it’s well-formed; provide some examples and motivate intuition; if there’s time, do preliminary empirical validation. Send this to a workshop to get feedback on the idea. Also, keep in mind that workshops are meant for preliminary research, not preliminary papers. A workshop paper still has to be a complete, well-written paper.
- If the idea looks like a good one, empirically validate it and firm up the theory. Send this to a good conference. [In computer science, RPT is based on good conferences rather than good journals.]
- If enthusiasm for the idea is high, write the journal article a year or so later, when you’ve had time to distill the essence and the impact of the work.
To achieve an iterative work-flow, make iterations easy:
- Once you know you’re going to do something, start on it right away: create a blank document file, create a blank presentation file, start drafting the email (with To: field blank). Then, if at any point in the future, you’re moved to work on it, the transaction cost of doing a little more work is near-zero.
- Work on a project whenever you’re moved to work on it. Don’t pay attention to deadline ordering unless it’s an n-day project, and only n free days are left.
The conclusion of this paper is that it is best to use an asymmetric alpha and a symmetric beta. I love its interpretation. “An asymmetric Dirichlet prior over theta serves to share commonalities across documents and a symmetric Dirichlet prior over phi serves to avoid conflicts between topics.”
Meng Wang, Bo Liu, Linjun Yang, Xian-Sheng Hua
- Accessibility-based reranking
- Image recoloring (for better discrimination)
- Color indication (tooltip)
- Is it really good to discriminate colors? Isn’t it different from how the colorblind perceive?
처음으로 작성해본 논문이라는 데에 의의.
How to do Research?
How to write an A+ research paper?
How to get your SIGGRAPH paper rejected?
How to live a Happy and Rewarding Life?
How to be a successful student?
How to Get a PhD?
Excerpted from Yuwing’s slide