Comment on "Language Trees and Zipping"

This is the extended version of a Comment submitted to Physical Review Letters. I first point out the inappropriateness of publishing a Letter unrelated to physics. Next, I give experimental results showing that the technique used in the Letter is 3 times worse and 17 times slower than a simple baseline. And finally, I review the literature, showing that the ideas of the Letter are not novel. I conclude by suggesting that Physical Review Letters should not publish Letters unrelated to physics. A recent Letter to Physical Review Letters, “Language Trees and Zipping,” by Benedetto et al. (2002) (available at http://arXiv.org/abs/cond-mat/ 0108530) is flawed in several ways. First of all, the Letter had nothing to do with physics, and instead belonged in a computer science journal, if it deserved to be published at all. Second of all, the actual results are unimpressive: as I will show, the techniques used lead to 3 times as many errors and are 17 times slower than a very simple baseline model applied to a standard, similar problem. Finally, the ideas in the Letter are not even novel: they are well known to those in several areas of computer science. The actual paper is clearly unrelated to physics, and much more closely related to areas of computer science such as Computational Linguistics and Machine Learning, as can be seen simply by reading the abstract of their paper, which I include here: In this Letter we present a very general method for extracting information from a generic string of characters, e.g., a text, a DNA sequence, or a time series. Based on data-compression techniques, its key point is the computation of a suitable measure of the remoteness of two bodies of knowledge. We present the implementation of the method to linguistic motivated problems, featuring highly accurate results for language recognition, authorship attribution, and language classification.