>I object to the implication that because this is found through the
algorithm that necessarily >shows some sort of bias on the part of
linguiststs.
There is an assumption, based on actual experience, that any input data is
likely to be skewed. Choices have to be made about which languages, which
vocabulary items, and which features are used. Even if the algorithm is
excellent, different choices will lead to different results. So perhaps it
isn't "bias" so much as recognition of the complexity and the nature of the
judgments needed.
Peter