Porting a Stochastic Part-of-Speech Tagger to Swedish

A b stract The Xerox Part-of-Speech Tagger (XPOST) claims to be practical. One aspect of practicality as defined here is reusability. Thus it is meant to be easy to port XPOST to a new language. To test this, XPOST was ported to Swedish. This port is described and evaluated. In previous work on part-of-speech tagging, a practical part-of-speech tagger was defined as one with the following set of properties (Cutting et a /1992): 1 • accu rate A tagger should assig n the co rre c t p a rt of det/2 n modal v det adj/2 n/2 prep speech to every word in the n prep/2 det n prep/4 det t e x t. n While 100% accuracy is desirable, it may not in fact be achievable. When text is manually tagged by several linguists, the tags assigned differ by a few percent, suggesting an effective upper-bound for tagging accuracy (Church 1989). fa st Ideally, the addition of part-of-speech tagging to a system will not significantly alter the speed with which text is processed. This may be difficult to evaluate, as systems which incorporate tagging may not operate at all without tagging. As a surrogate, one may compare the cost of assigning tags with that of simply extracting words from text— tokenization. If tagging is not significantly slower than tokenizing then its performance impact on complex text processing systems should certainly be minimal.