Using a Morphological Database to Increase the Accuracy in POS Tagging

Proceedings of Recent Advances in Natural Language Processing, pages 49–55, Hissar, Bulgaria, 12-14 September 2011. Using a Morphological Database to Increase the Accuracy in POS Tagging Hrafn Loftsson School of Computer Science Reykjavik University hrafn@ru.is Sigrún Helgadóttir The Árni Magnússon Institute for Icelandic Studies sigruhel@hi.is Eiríkur Rögnvaldsson Department of Icelandic University of Iceland eirikur@hi.is Abstract We experiment with extending the dic- tionaries used by three open-source part- of-speech taggers, by using data from a large Icelandic morphological database. We show that the accuracy of the tag- gers can be improved significantly by us- ing the database. The reason is that the unknown word ratio reduces dramatically when adding data from the database to the taggers’ dictionaries. For the best per- forming tagger, the overall tagging accu- racy increases from the base tagging result of 92.73% to 93.32%, when the unknown word ratio decreases from 6.8% to 1.1%. When we add reliable frequency informa- tion to the tag profiles for some of the words originating from the database, we are able to increase the accuracy further to 93.48% – this is equivalent to 10.3% error reduction compared to the base tagger. 1 Introduction In general, part-of-speech (PoS) taggers can be catagorised into two types. First, data-driven tag- gers, i.e. taggers that are trained on pre-tagged corpora and are both language and tagset indepen- dent, e.g. (Brants, 2000; Toutanova et al., 2003; Shen et al., 2007). Second, linguistic rule-based taggers, which are developed “by hand” using lin- guistic knowledge, with the purpose of tagging a specific language using a particular tagset, e.g. (Karlsson et al., 1995; Loftsson, 2008). All taggers use a particular tagset T and rely on a dictionary D containing the tag profile (am- biguity class) Tw for each word w. A tag profile Tw indicates which tags are assignable to w, thus Tw ⊂ T. Essentially, for each word w, a tagger disambiguates Tw by selecting (or removing all but) one tag from it with regard to context. The dictionary D is derived by a data-driven tagger during training, and derived or built during devel- opment of a linguistic rule-based tagger. When tagging new text, PoS taggers frequently encounter words that are not in D, i.e. so-called unknown words. An unknown word u can be quite problematic for a tagger, because the tag profile for u needs to be guessed. In most cases, PoS tag- gers therefore contain a special module, called an unknown word guesser, to generate the tag profile for unknown words. Frequently, the guessing of the tag profile for unknown words is incorrect and therefore the tagging accuracy for these words is considerably lower than the tagging accuracy for known words. To increase the overall tagging ac- curacy of PoS taggers, one might therefore try to refine the underlying unknown word guessers. An- other approach is simply to try to minimise the ra- tio of unknown words by extending the dictionar- ies used by the taggers. In this paper, we use the latter approach. We ex- …

Content truncated to 500 words