Machine Learning (Theory)


Six Months

Tags: General jl@ 10:03 pm

This is the 6 month point in the “run a research blog” experiment, so it seems like a good point to take stock and assess.

One fundamental question is: “Is it worth it?” The idea of running a research blog will never become widely popular and useful unless it actually aids research. On the negative side, composing ideas for a post and maintaining a blog takes a significant amount of time. On the positive side, the process might yield better research because there is an opportunity for better, faster feedback implying better, faster thinking.

My answer at the moment is a provisional “yes”. Running the blog has been incidentally helpful in several ways:

  1. It is sometimes educational. example
  2. More often, the process of composing thoughts well enough to post simply aids thinking. This has resulted in a couple solutions to problems of interest (and perhaps more over time). If you really want to solve a problem, letting the world know is helpful. This isn’t necessarily because the world will help you solve it, but it’s helpful nevertheless.
  3. In addition, posts by others have helped frame thinking about “What are important problems people care about?”, and why. In the end, working on the right problem is invaluable.
9 Comments to “Six Months”
  1. Maverick Woo says:

    Yes, this blog is totally worth it.

    You may want to post a bit slower though. It certainly takes a lot of time to write your typical posts since, unlike in a personal blog, you actually have something concrete that you want to say. And your audience are also mostly overloaded academics and they may not be able to react to something that becomes “obsolete” in a week.

  2. Arindam Banerjee says:

    I think the blog is a very good idea and you are doing an excellent job. I am sure many people regularly read your blog, although very few actually post (do you have any stats?).

    I would like to see more posts on `I liked reading this paper’. Since the number of papers grow by the day, its useful to collaboratively filter out the good ones.

  3. The number of unique visitors per day is something between 100 and 200.

    I’ll try emphasize ‘good paper’ posts a bit more. If you see one, feel free to run it by me.

  4. Hal Daume says:

    I agree with Arindam; in fact, it might be useful to have a separate section for this, where people can post selected good papers from recent conferences in their own field.

    ACL just completed, so to get things rolling, here’s my personal, biased, list of cool papers from this conference (see for online papers):

    David Chiang, A Hierarchical Phrase-Based Model for Statistical Machine Translation. (Best paper award.) This paper takes the standard phrase-based MT model that is popular in our field (basically, translate a sentence by individually translating phrases and reordering them according to a complicated statistical model) and extends it to take into account hierarchy in phrases, so that you can learn things like “X ‘s Y” -> “Y de X” in chinese, where X and Y are arbitrary phrases. This takes a step toward linguistic syntax for MT, which our group is working strongly on, but doesn’t require any linguists to sit down and write out grammars or parse sentences.

    Rie Kubota Ando and Tong Zhang, A High-Performance Semi-Supervised Learning Method for Text Chunking. This is more of a machine learning style paper, where they improve a sequence labeling task by augmenting it with models from related tasks for which data is free. I.e., I might train a model that, given a context with a missing word, will predict the word (eg., “The ____ gave a speech” might want you to insert “president”.) By doing so, you can use these other models to give additional useful information to your main task.

    Noah A. Smith and Jason Eisner, Contrastive Estimation: Training Log-Linear Models on Unlabeled Data. This paper talks about training sequence labeling models in an unsupervised fashion, basically by contrasting what the model does on the correct string with what the model does on a corrupted version of the string. They get significantly better results than just by using EM in an HMM, and the idea is pretty nice.

    Patrick Pantel, Inducing Ontological Co-occurrence Vectors. This is a pretty neat idea (though I’m biased — Patrick is a friend) where one attempts to come up with feature vectors that describe nodes in a semantic hierarchy (ontology) that could enable you to figure out where to insert new words that are not in your ontology. The results are pretty good, and the method is fairly simple; I’d imagine that a more complex model/learning framework could improve the model even further.

  5. I certainly find it educational to see how other people think about what they view as the important problems.

    Academics are usually secretive about their unpublished ideas to some degree. Do you have any concerns about “leaking” ideas via the blog that others go on to publish?

    Of course, someone should reference or acknowledge the blog if it was influential in their work.

  6. I am not very concerned about ‘idea stealing’. One basic component of this is an excess of ideas and deficit of time on my part.

    Another component is that I have only a mild preference over solving a problem personally versus simply having it solved.

    Another is that the idea of ‘idea stealing’ is well-established in academia and there are mechanisms to cope with it. An idea is not stolen if a citation to prior work is included. Hopefully, people inspired by the blog will cite it in an appropriate manner.

    One thing which does worry me is duplication of work. Hopefully, the site itself will act as a coordination point to avoid this.

  7. Johnny Cui says:

    Hi,John. I’m glad to have found this excellent site. I’m a postgraduate major in NLP and IR in China. Your blog really benefits me a lot. In fact I have the same feeling about maintaining a dedicated blog for research experience although we have different spoken language. As we saw in the comments, I’m confident in running such a blog.

    I do not need large pageview everyday. What I really care is the group of people who have the same interested area as me. There is an old saying in Chinese that: “Articles are clusterd according to their genus, man are grouped according to their community”. That’s true.

  8. I say this because, …

    … inspired by this guy, I’ve been thinking it might be useful to start a research blog.

  9. Omid Madani says:

    I have enjoyed and certainly benefited from your blog and others’ contributions. Thanks John!

Leave a Reply

Powered by WordPress