An Unsupervised Morpheme-Based HMM for Hebrew Morphological Disambiguation

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 665–672, Sydney, July 2006. c⃝2006 Association for Computational Linguistics An Unsupervised Morpheme-Based HMM for Hebrew Morphological Disambiguation Meni Adler Department of Computer Science Ben Gurion University of the Negev 84105 Beer Sheva, Israel adlerm@cs.bgu.ac.il Michael Elhadad Department of Computer Science Ben Gurion University of the Negev 84105 Beer Sheva, Israel elhadad@cs.bgu.ac.il Abstract Morphological disambiguation is the pro- cess of assigning one set of morphologi- cal features to each individual word in a text. When the word is ambiguous (there are several possible analyses for the word), a disambiguation procedure based on the word context must be applied. This paper deals with morphological disambiguation of the Hebrew language, which combines morphemes into a word in both agglutina- tive and fusional ways. We present an un- supervised stochastic model – the only re- source we use is a morphological analyzer – which deals with the data sparseness prob- lem caused by the affixational morphology of the Hebrew language. We present a text encoding method for languages with affixational morphology in which the knowledge of word formation rules (which are quite restricted in He- brew) helps in the disambiguation. We adapt HMM algorithms for learning and searching this text representation, in such a way that segmentation and tagging can be learned in parallel in one step. Results on a large scale evaluation indicate that this learning improves disambiguation for complex tag sets. Our method is applicable to other languages with affix morphology. 1 Introduction Morphological disambiguation is the process of as- signing one set of morphological features to each individual word in a text, according to the word context. In this work, we investigate morphological dis- ambiguation in Modern Hebrew. We explore unsu- pervised learning method, which is more challeng- ing than the supervised case. The main motivation for this approach is that despite the development ∗This work is supported by the Lynn and William Frankel Center for Computer Sciences, and by the Knowledge Center for Hebrew Processing, Israel Sci- ence Ministry. of annotated corpora in Hebrew1, there is still not enough data available for supervised training. The other reason, is that unsupervised methods can handle the dynamic nature of Modern Hebrew, as it evolves over time. In the case of English, because morphology is simpler, morphological disambiguation is generally covered under the task of part-of-speech tagging. The main morphological variations are embedded in the tag name (for example, Ns and Np for noun singular or plural). The tagging accuracy of supervised stochastic taggers is around 96%- 97% (Manning and Schutze, 1999, 10.6.1). Meri- aldo (1994) reports an accuracy of 86.6% for an un- supervised word-based HMM, trained on a corpus of 42,186 sentences (about 1M words), over a tag set of 159 different tags. Elworthy (1994), in con- trast, reports an accuracy of 75.49%, 80.87% and 79.12% for unsupervised word-based HMM trained on parts of the LOB corpora, with a tagset of 134 tags. With …

Content truncated to 500 words