Fintan - Flexible, Integrated Transformation and Annotation eNgineering

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 7212–7221 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 7212 Fintan - Flexible, Integrated Transformation and Annotation eNgineering Christian F¨ath, Christian Chiarcos, Bj¨orn Ebbrecht, Maxim Ionov Applied Computational Linguistics Lab (ACoLi) Goethe University Frankfurt, Germany {faeth, chiarcos, ebbrecht, ionov}@informatik.uni-frankfurt.de Abstract We introduce the Flexible and Integrated Transformation and Annotation eNgeneering (Fintan) platform for converting heterogeneous linguistic resources to RDF. With its modular architecture, workflow management and visualization features, Fintan facilitates the development of complex transformation pipelines by integrating generic RDF converters and augmenting them with extended graph processing capabilities: Existing converters can be easily deployed to the system by means of an ontological data structure which renders their properties and the dependencies between transformation steps. Development of subsequent graph transformation steps for resource transformation, annotation engineering or entity linking is further facilitated by a novel visual rendering of SPARQL queries. A graphical workflow manager allows to easily manage the converter modules and combine them to new transformation pipelines. Employing the stream-based graph processing approach first implemented with CoNLL-RDF, we address common challenges and scalability issues when transforming resources and showcase the performance of Fintan by means of a purely graph-based transformation of the Universal Morphology data to RDF. Keywords: data formats, interoperability, RDF, graphs, linguistic annotation, tab-separated values, universal morphology, CoNLL-RDF 1. Motivation In recent years, the focus in the development of Nat- ural Language Processing (NLP) applications has con- stantly shifted from algorithmic, rule-based strategies to- wards more data-driven approaches. This includes not only statistical methods but also neural networks and deep learning. However, all these methods heavily rely on the amount of available language resources and their data struc- tures. The EU-funded Prˆet-`a-LLOD project1 aims at tack- ling the crucial challenges of discovering, linking and trans- forming language resources and making them available as interoperable Linked Data (LD) using well-established RDF-based2 formats like OntoLex-Lemon (Cimiano et al., 2016), CoNLL-RDF3 (Chiarcos and F¨ath, 2017; Chiarcos and Schenk, 2018) or the NLP Interchance Format (Hell- mann et al., 2012, NIF). While resource discovery, maintenance and licensing is covered by other parts of the project, in this paper we focus on the Flexible and Integrated Transformation and Annotation eNgineering (Fintan) platform which pro- vides a workflow-driven, modular approach to graph- transformation while still enabling the usage of well- established resource-specific converters. 2. Challenges in Linguistic Linked Data 2.1. Heterogeneity of Linguistic Resources Since language resources on the web are highly heteroge- neous and tailored towards specific use cases, they most 1https://www.pret-a-llod.eu 2The Resource Description Framework is the W3C standard for rendering data in the Semantic Web (McBride, 2004). 3CoNLL-RDF is an isomorphic rendering of the CoNLL for- mat family in RDF. CoNLL formats are task-specific TSV rep- resentations used in CoNLL shared tasks which have become a de-facto standard for NLP tools developed in their context, cf. http:/www.signll.org/conll often cannot easily be adapted to other NLP applications beyond their intended scope. Corpora may …

Content truncated to 500 words