Named Entity Resources - Overview and Outlook

Named Entity Resources - Overview and Outlook Maud Ehrmann1, Damien Nouvel2, Sophie Rosset3 1 Digital Humanities Laboratory, EPFL, Switzerland 2 ERTIM, Inalco, Paris, France 3 LIMSI, CNRS, Université Paris-Saclay, F-91405 Orsay, France maud.ehrmann@epfl.ch, damien.nouvel@inalco.fr, sophie.rosset@limsi.fr Abstract Recognition of real-world entities is crucial for most NLP applications. Since its introduction some twenty years ago, named entity processing has undergone a significant evolution with, among others, the definition of new tasks (e.g. entity linking) and the emergence of new types of data (e.g. speech transcriptions, micro-blogging). These pose certainly new challenges which affect not only methods and algorithms but especially linguistic resources. Where do we stand with respect to named entity resources? This paper aims at providing a systematic overview of named entity resources, accounting for qualities such as multilingualism, dynamicity and interoperability, and to identify shortfalls in order to guide future developments. Keywords: named entity, linguistic resources, NE typologies, annotated corpora, evaluation, linked data 1. Introduction Recognition of real-world entities is crucial for most, if not all, text mining applications. Indeed, referential units such as name of persons, places and organisations underlie the semantics of texts and guide their interpretation. Acknowl- edged some twenty years ago, named entity (NE) mining is an operation of ever increasing importance for many NLP applications which has undergone major evolutions since then. Named entity processing is representative of the evolution of information extraction from a document to a semantic- centric view point (Rao et al., 2013). As first introduced during the 6th Message Understanding Conference (Grish- man and Sundheim, 1995), it corresponds to the recognition and classification of entities of interest in texts, generally of type Person, Organisation and Location. During sub- sequent evaluation campaigns and research programs, the task quickly broadened and became more complex, with the extension and refinement of typologies (Sekine et al., 2002; Galibert et al., 2011), the diversification of languages taken into account (Tjong Kim Sang and De Meulder, 2003; San- tos et al., 2006; Magnini et al., 2008a; Galliano et al., 2009; Benikova et al., 2014), and the expansion of the linguistic scope with, along proper names, the consideration of nom- inal phrases and pronouns as candidate lexical units (Dod- dington et al., 2004). Later on, as recognition and classi- fication were reaching satisfying performances (at least for well-covered languages and main stream texts such as news articles), attention focused on finer-grained processing, e.g. metonymy recognition (Markert and Nissim, 2009), and on the next logical step, namely disambiguation. Entity resolution has first been defined as a clustering prob- lem, where different mentions in texts referring to the same entity must be grouped together (Mann and Yarowsky, 2003; Artiles et al., 2008). Next, with the advent of knowledge bases (KB) containing plenty of entities along with detailed information (Hovy et al., 2013), entity dis- ambiguation switched from clustering mentions to align- ing mentions to unique identifiers in a KB. Thereupon, the task of entity linking gained strong impetus (Ji and Gr- ishman, 2011; Shen et al., 2015) and is now …

Content truncated to 500 words