Just how far can concept searching and semantic analysis go in automating document review in discovery?
No one knows yet. The answer depends in part on another question: Is the ability of software accurately to assess responsive documents in a large collection constrained by linguistics (algorithms) or by computational limits (the cost of raw processing power)?
Advanced e-discovery software has proliferated recently. My sense is that these products are limited by the underlying techniques of full-text search and analysis, not by limits on affordable processing power. (See my July 2004 post, Developments in Full Text Searching, in which I share comments of a computational linguist on this issue).
If, however, the limit is processing data in a reasonable amount of time, then advances in computing speed would improve results. A recent article suggests that the moral equivalent of supercomputing will soon be available to the mass market. Google and the Wisdom of Clouds in the December 24th issue of Business Week explains that Google and IBM are developing “cloud computing,” the ability to use the vast networks of servers as a virtual supercomputer. If this succeeds, then lawyers will be able to run search software “in the cloud.” (I am assuming confidentiality issues can be resolved.) At that point, we would know better just how much massive computing power can replace human review.