9/1/09

Grounded Neural Networking

If you are into neural nets, you know about the Laboratory of Computer and Information Science at the Helsinki University of Technology. One of the Department's most important professors is Teuvo Kohonen, the creator of the self-organizing map algorithm (SOM). The deparment also provides one of the best shareware downloads (SOM Toolbox) for using the SOM--it runs in the MatLab environment.

As I discussed in a previous blog (4/01/09), in 2003 I published an article in Symbolic Interaction exploring how qualitative researchers can use the SOM to conduct grounded theoretical investigations of large, complex, numerical databases. For the next six years, I sat around hoping someone other than myself would find the idea interesting and useful. Nothing happened! I know that publishing on mixed methods seldom goes anywhere, but I thought that, with the incredible advances taking place in complexity science and informatics and the internet, qualitative researchers would eventually consider the idea.

They have yet to do so. But, perhaps the latest article by Nina Janasik, Timo Honkela, and Henrik Bruun of the Helsinki University of Technology can change people's minds.

The title of their article is TEXT MINING IN QUALITATIVE RESEARCH. The purpose of the article is to show qualitative researchers how to apply the SOM to qualitative data.


Here is the abstract:

---------------------------------------------
ABSTRACT
The article provides an introduction to and a demonstration of the self-organizing map (SOM) method for organizational researchers interested in the use of qualitative data. The SOM is a versatile quantitative method very commonly used across many disciplines to analyze large data sets. The outcome of the SOM analysis is a map in which entities are positioned according to similarity. The authors' argument is that text mining using the SOM is particularly effective in improving inference quality within qualitative research. SOM creates multiple well-grounded perspectives on the data and thus improves the quality of the concepts and categories used in the analysis.
---------------------------------------------

When I read this article I was more than a little excited! I cannot tell you how much time I spent between 2001 and 2003 at the Helsinki website trying to learn about the SOM and figuring out how to use the SOM Toolbox. Let's just say it was a lot and leave it at that. I also cannot tell you how much respect I have for the researchers there. Incredible research; they make their work and software freely available to others. It is just fantastic.

I also have to say that Janasik, Honkela, and Bruun do an excellent job addressing the limitations of my 2003 article--in particular, how I did not go far enough in demonstrating just how useful the SOM is for qualitative work. As such, I agree completely with their critique. And, I agree that any qualitative researcher trying to figure out how to do their work in the digital age should read this article.

Enough said...





Key Words: grounded theory • constructivism • self-organizing map • text mining • document interpretation

9 comments:

  1. Hi
    That was amazing for me to first find your 2003 paper and then your post here. :)
    I am a PhD student at ETHZ, with background in computer science and modeling and now doing research in the field of Architecture and Urban Design.
    I am deeply interested in SOM, not just as a data mining tool, but as an environment for doing a lot of things, and just today, came across the idea of Grounded Theory, and found out that the "concepts" there should be mathematically equal to eigen-vector of observations (clusters) and then since we have a vectorial representation, we could easily use SOM for modeling.
    Then it could be a process from "rough observations or a lot of descriptions to a quantitative design and modeling (e.g. in Urban design problems with people expressions of a place), and this approach is potentially so different than existing "reductionistic and limited" models of the socio-economic environments.

    Then, I found your post here, and just wanted to say a Hello and say that after 9 years, one person is interested in this cross disciplinary use of SOM and GT.

    Best
    Vahid

    ReplyDelete
    Replies
    1. Hi Vahid

      Thanks so much for the posting. I am glad the book and article are useful. I have other articles, if you are interested, at my academic website:
      http://www.personal.kent.edu/~bcastel3/SACS-Toolkit.html

      In fact, we just finished a mathematical outline of our approach to case-based computational modeling using the SOM in combination with a set of other techniques.

      Delete
  2. Hi
    Thanks for sharing more information.
    This combination sounds very interesting, while I assume it could be somehow a dynamic SOM for classification with dynamic labels.
    Currently, I am developing a new SOM, in which a neuron is an index for connecting two independent, but interactive spaces, which one is informational and the other is physical space, while two spaces are dynamic. I think it could be an instance of cartesian dualism.
    If you were interested, I would like to share my results with you.
    Best
    Vahid

    ReplyDelete
    Replies
    1. This is very interesting. I am interested in being in the loop and could share with you (Vahid and Brian) a couple recent articles.

      Delete
    2. Hi
      Timo
      That is very nice you are here. I couldn't expect it. :)
      Because, exactly I have open your papers in my computer these days.
      These days, I am reading and thinking about texts as qualitative data without any specific assumptions for representation of the objects of interests. and then, we can find your groups' projects there, including websom.
      But, for the case of this dynamic SOM, which I mentioned I think it could be a different interpretation for SOM, with some more constraints relaxation.
      Best
      Vahid

      Delete
  3. that would be very cool. email me at my academic address: bcastel3 (at) kent (dot) edu

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. Hi
      Thanks, sure.
      That is my honor to share my ideas with you.
      further, I am not sure since I was sleepy last night. I think I had a very long comment about your SACS toolkit.

      Best
      Vahid

      Delete