Joint Lemmatization and Morphological Tagging with Lemming

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2268–2274, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Joint Lemmatization and Morphological Tagging with LEMMING Thomas M¨uller1 Center for Information and Language Processing1 University of Munich, Germany muellets@cis.lmu.de Ryan Cotterell1,2 Alexander Fraser1 Department of Computer Science2 Johns Hopkins University, USA ryan.cotterell@jhu.edu Hinrich Sch¨utze1 Abstract We present LEMMING, a modular log- linear model that jointly models lemmati- zation and tagging and supports the inte- gration of arbitrary global features. It is trainable on corpora annotated with gold standard tags and lemmata and does not rely on morphological dictionaries or an- alyzers. LEMMING sets the new state of the art in token-based statistical lemmati- zation on six languages; e.g., for Czech lemmatization, we reduce the error by 60%, from 4.05 to 1.58. We also give em- pirical evidence that jointly modeling mor- phological tags and lemmata is mutually beneficial. 1 Introduction Lemmatization is important for many NLP tasks, including parsing (Bj¨orkelund et al., 2010; Seddah et al., 2010) and machine translation (Fraser et al., 2012). Lemmata are required whenever we want to map words to lexical resources and establish the relation between inflected forms, particularly crit- ical for morphologically rich languages to address the sparsity of unlemmatized forms. This strongly motivates work on language-independent token- based lemmatization, but until now there has been little work (Chrupała et al., 2008). Many regular transformations can be described by simple replacement rules, but lemmatization of unknown words requires more than this. For instance the Spanish paradigms for verbs end- ing in ir and er share the same 3rd person plu- ral ending en; this makes it hard to decide which paradigm a form belongs to.1 Solving these kinds of problems requires global features on the lemma. Global features of this kind were not supported 1Compare admiten “they admit” → admitir “to admit”, but deben “they must” → deber “to must”. by previous work (Dreyer et al., 2008; Chrupała, 2006; Toutanova and Cherry, 2009; Cotterell et al., 2014). There is a strong mutual dependency between (i) lemmatization of a form in context and (ii) disambiguating its part-of-speech (POS) and mor- phological attributes. Attributes often disam- biguate the lemma of a form, which explains why many NLP systems (Manning et al., 2014; Padr´o and Stanilovsky, 2012) apply a pipeline approach of tagging followed by lemmatization. Conversely, knowing the lemma of a form is of- ten beneficial for tagging, for instance in the pres- ence of syncretism; e.g., since German plural noun phrases do not mark gender, it is important to know the lemma (singular form) to correctly tag gender on the noun. We make the following contributions. (i) We present the first joint log-linear model of mor- phological analysis and lemmatization that oper- ates at the token level and is also able to lem- matize unknown forms; and release it as open- source (http://cistern.cis.lmu.de/lemming). It is trainable on corpora annotated with gold stan- dard tags and lemmata. Unlike other work (e.g., …

Content truncated to 500 words