Combining Hand-crafted Rules and Unsupervised Learning in Constraint-based Morphological Disambiguation

This paper presents a constraint-based morphological disambiguation approach that is applicable languages with complex morphology-specifically agglutinative languages with productive inflectional and derivational morphological phenomena. In certain respects, our approach has been motivated by Brill's recent work (Brill, 1995b), but with the observation that his transformational approach is not directly applicable to languages like Turkish. Our system combines corpus independent handcrafted constraint rules, constraint rules that are learned via unsupervised learning from a training corpus, and additional statistical information from the corpus to be morphologically disambiguated. The hand-crafted rules are linguistically motivated and tuned to improve precision without sacrificing recall. The unsupervised learning process produces two sets of rules: (i) choose rules which choose morphological parses of a lexical item satisfying constraint effectively discarding other parses, and (ii) delete rules, which delete parses satisfying a constraint. Our approach also uses a novel approach to unknown word processing by employing a secondary morphological processor which recovers any relevant inflectional and derivational information from a lexieal item whose root is unknown. With this approach, well below 1% of the tokens remains as unknown in the texts we have experimented with. Our results indicate that by combining these hand-crafted, statistical and learned information sources, we can attain a recall of 96 to 97% with a corresponding precision of 93 to 94%, and ambiguity of 1.02 to 1.03 parses per token. Automatic morphological disambiguation is a very crucial component in higher level analysis of natural language text corpora. Morphological disambiguation facilitates parsing, essentially by performing a certain amount of ambiguity resolution using relatively cheaper methods (e.g., Gfing6rdii and Oflazer (1995)). There has been a large number of studies in tagging and morphological disambiguation using various techniques. Part-of-speech tagging systems have used either a statistical approach where a large corpora has been used to train a probabilistic model which then has been used to tag new text, assigning the most likely tag for a given word in a given context (e.g., Church (1988), Cutting et al. (1992), DeRose (1988)). Another approach is the rule-based or constraint-based approach, recently most prominently exemplified by the Constraint Grammar work (Karlsson et al., 1995; Voutilainen, 1995b; Voutilainen et al., 1992; Voutilainen and Tapanainen, 1993), where a large number of hand-crafted linguistic constraints are used to eliminate impossible tags or morphological parses for a given word in a given context. Brill (1992; 1994; 1995a) has presented a transformation-based learning approach, which induces rules from tagged corpora. Recently he has extended this work so that learning can proceed in an unsupervised manner using an untagged corpus (Brill, 1995b). Levinger et al. (1995) have recently reported on an approach that learns morpholexical probabilities from untagged corpus and have the used the resulting information in morphological disambiguation in Hebrew. In contrast to languages like English, for which there is a very small number of possible word forms with a given root word, and a small number of tags associated with a given lexical form, languages like Turkish or Finnish with very productive agglutinative morphology where it is possible to produce thousands of forms (or even millions (Hankamer, 1989)) for a given root word, pose a challenging problem for morphological disambiguation. In English, for example, a word such as make or set can be verb