The Morpho-syntactic Annotation of Animacy for a Dependency Parser

The Morpho-syntactic Annotation of Animacy for a Dependency Parser Mohammed Attia, Vitaly Nikolaev, Ali Elkahky Google Inc. New York, USA {attia,vitalyn,alielkahky}@google.com Abstract In this paper we present the annotation scheme and parser results of the animacy feature in Russian and Arabic, two morphologically- rich languages, in the spirit of the universal dependency framework (McDonald et al., 2013; de Marneffe et al., 2014). We explain the animacy hierarchies in both languages and make the case for the existence of five animacy types. We train a morphologi- cal analyzer on the annotated data and the results show a prediction f-measure for animacy of 95.39% for Russian and 92.71% for Arabic. We also use animacy along with other morphological tags as features to train a dependency parser, and the results show a slight improvement gained from animacy. We compare the impact of animacy on improving the dependency parser to other features found in nouns, namely, ‘gender’, ‘number’, and ‘case’. To our knowledge this is the first contrastive study of the impact of morphological features on the accuracy of a transition parser. A portion of our data (1,000 sentences for Arabic and Russian each, along with other languages) annotated according to the scheme described in this paper is made publicly available (https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-1983) as part of the CoNLL 2017 Shared Task on Multilingual Parsing (Zeman et al., 2017). Keywords: animacy, Arabic, Russian, Dependency Parsing 1. Introduction The explicit encoding and identification of the animacy of entities in data has been reported to improve NLP tasks such as syntactic function disambiguation (Øvrelid, 2004; Lamers, 2007), anaphora resolution (Orasan and Evans, 2007a; Liang and Wu, 2004; Singh et al., 2014), syntac- tic parsing (Marton et al., 2011; Ambati et al., 2010; Nivre et al., 2008; Bharati et al., 2008), and disambiguation ac- curacy in boosting fluency in text generation (Bloem and Bouma, 2013). In human cognition research, animacy is considered as one of the first characterizations made by human beings in their infancy and the last distinction lost in adults with Alzheimer’s disease (Szewczyk and Schriefers, 2010). This is probably why computational representation of languages (particularly where animacy plays a morpho-syntactic func- tion, e.g. Russian, Arabic, and Hindi) needs to provide ad- equate description and annotation of this feature. Contrary to previously prevalent perception that animacy is merely a semantic feature, recent years have seen increasing research showing animacy as permeating all levels of linguistic rep- resentations: morphology, syntax, semantics, and discourse (Folli and Harley, 2008; Ramchand, 2008; Ritter, 2014; Szewczyk and Schriefers, 2010). Moreover, animacy anno- tation for some languages is important for both downstream tasks as well as generation. Animacy is a linguistic property that impacts and relates to a number of other linguistic phenomena, such as case marking, agreement, topicality, argument realization, and structural preferences. It has received formal linguistic in- vestigation in the works of various researchers, most no- tably (Silverstein, 1976; Dik, 1997; Aissen, 2003; Martin et al., 2005; Dahl, 2008; Bloem and Bouma, 2013; Eck- hoff, 2015; Karsdorp et al., 2015). …

Content truncated to 500 words