none 10.1075/nlp.14.10joh John Benjamins Publishing Company John Benjamins Publishing Company 1757 131903026 3924503 1420520 202507030024000027 10.1075 2025-07-02T18:24:45Z 2021-11-09T09:12:16Z 1 1567-8202 10.1075/nlp https://benjamins.com/ catalog/nlp Dana Dannélls University of Gothenburg https://ror.org/01tm6cn81 Lars Borin University of Gothenburg https://ror.org/01tm6cn81 Karin Friberg Heppin HeppiLing AB Harmonization, integration, method development and practical language technology applications Large computational lexicons are central NLP resources. Swedish FrameNet++ aims to be a versatile full-scale lexical resource for NLP containing many kinds of linguistic information. Although focused on Swedish, this ongoing effort, which includes building a new Swedish framenet and recycling existing lexicons, has offered valuable insights into general aspects of lexical-resource building for NLP, which are discussed in this book: computational and linguistic problems of lexical semantics and lexical typology, the nature of lexical items (words and multiword expressions), achieving interoperability among heterogeneous lexical content, NLP methods for extending and interlinking existing lexicons, and deploying the new resource in practical NLP applications. This book is targeted at everyone with an interest in lexicography, computational lexicography, lexical typology, lexical semantics, linguistics, computational linguistics and related fields. We believe it should be of particular interest to those who are or have been involved in language resource creation, development and evaluation. 14 11 26 2021 11 9 2021 9789027209900 9789027258489 John Benjamins Publishing Company Amsterdam 10.1075/nlp.14 http://www.jbe-platform.com/content/books/9789027258489 https://www.jbe-platform.com/deliver/fulltext/9789027258489.pdf https://www.degruyter.com/document/doi/10.1075/nlp.14/html Richard Johansson University of Gothenburg https://ror.org/01tm6cn81 Karin Friberg Heppin HeppiLing AB Dimitrios Kokkinakis University of Gothenburg https://ror.org/01tm6cn81 Abstract We investigate the feasibility of automatic semantic role labeling (SRL) using Swedish FrameNet (SweFN). In the first part of the chapter, we describe a baseline system using a traditional division into segmentation and labeling steps. These subsystems are implemented as separate machine learning models, and we explore a wide range of syntactic and lexical features for these models. In the second part, we turn to the question of how the frame-to-frame relations defined in FrameNet allow us to use the annotated examples more effectively. The cross-frame generalization methods reduce the number of errors made by the labeling classifier by 27%. For previously unseen frames, the reduction is even more significant: 50%. 11 26 2021 11 9 2021 263 280 10.1075/nlp.14.10joh https://benjamins.com/catalog/nlp.14.10joh https://www.degruyter.com/document/doi/10.1075/nlp.14.10joh/html Proceedings of Coling 1998 Baker 86 1998 The Berkeley FrameNet project Proceedings of EURALEX 2010 Borin 269 2010 The past meets the present in Swedish FrameNet++ 10.1007/s10579-013-9233-4 Computational Linguistics Brown 18 4 467 1992 Class-based n-gram models of natural language Proceedings of ACL 2007 Collobert 560 2007 Fast semantic extraction using a novel neural network architecture Journal of Machine Learning Research Crammer 7 551 2006 Online passive-aggressive algorithms Proceedings of NAACL/HLT 2010 Das 948 2010 Probabilistic frame-semantic parsing 10.1353/lan.1991.0021 The linguistic annotation system of the Stockholm–Umeå corpus project – description and guidelines Ejerhed 1992 Journal of Machine Learning Research Fan 9 1871 2008 Liblinear: A library for large linear classification The Oxford handbook of linguistic analysis Fillmore 313 2009 A frames approach to semantic analysis Proceedings of LREC 2012 Friberg Heppin 256 2012 The rocky road towards a Swedish FrameNet – creating SweFN 10.3115/1609067.1609091 10.1162/089120102760275983 Proceedings of ACL 2002 Gildea 239 2002 The necessity of syntactic parsing for predicate argument recognition Proceedings of ACL 2007: Demo and poster sessions Halácsy 209 2007 HunPos – an open source trigram tagger Dependency-based semantic analysis of natural-language text Johansson 2008 10.3115/1699571.1699586 Proceedings of *SEM 2012 Johansson 95 2012 Non-atomic classification to improve a semantic role labeler for a low-resource language Proceedings of LREC 2012 Johansson 3697 2012 Semantic role labeling with the Swedish FrameNet 10.3115/1273073.1273130 10.3115/1621474.1621522 Proceedings of the Nodalida workshop FRAME 2007: Building frame semantics resources for Scandinavian and Baltic languages Johansson 27 2007 Using WordNet to extend FrameNet coverage 10.3115/1596324.1596355 10.3115/1599081.1599131 Proceedings of ACL/HLT 2008 Koo 595 2008 Simple semi-supervised dependency parsing 10.1162/coli.2008.34.2.145 10.3115/1687878.1687883 Proceedings of ACL 2012 Moschitti 759 2012 Modeling topic dependencies in hierarchical text categorization Treebanking for discourse and speech: Proceedings of the NODALIDA 2005 special session on treebanks for spoken language and discourse Nilsson 119 2006 MAMBA meets TIGER: Reconstructing a Swedish treebank from antiquity 10.1017/S1351324906004505 10.1162/0891201053630264 10.2200/S00239ED1V01Y200912HLT006 10.1109/ICDM.2009.129 Proceedings of EMNLP 2004 Xue 88 2004 Calibrating features for semantic role labeling