ABSTRACT
The first edition of this handbook appeared in 2000, but the project that resulted in that volume in fact began 4 years earlier, in mid-1996. When Hermann Moisl, Harold Somers, and I started planning the content of the book, the field of natural language processing was less than 10 years into what some might call its “statistical revolution.” It was still early enough that there were occasional signs of friction between some of the “old guard,” who hung on to the symbolic approaches to natural language processing that they had grown upwith, and the “young turks,” with their new-fangled statistical processing techniques, which just kept gaining ground. Some in the old guard would give talks pointing out that there were problems in natural languageprocessing thatwere beyond the reachof statistical or corpus-basedmethods; meanwhile, the occasional young turk could be heard muttering a variation on Fred Jelinek’s 1988 statement that “whenever I fire a linguist our system performance improves.” Then there were those with an eye to a future peaceful coexistence that promised jobs for all, arguing thatweneeded to develop hybrid techniques and applications that integrated the best properties of both the symbolic approaches and the statistical approaches.
