The Swedish FrameNet++ : : Harmonization, Integration, Method Development and Practical Language Technology Applications.

• Superior document: Natural Language Processing Series ; v.14
• Author: Dannélls, Dana.
• TeilnehmendeR: Borin, Lars.; Friberg Heppin, Karin.
• Place / Publishing House: Amsterdam/Philadelphia : : John Benjamins Publishing Company,, 2021.; {copy}2021.
• Year of Publication: 2021
• Edition: 1st ed.
• Language: English
• Series: Natural Language Processing Series
• Physical Description: 1 online resource (349 pages)

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.

Table of Contents

Intro -- The Swedish FrameNet++ -- Editorial page -- Title page -- Copyright page -- Table of contents -- Acronyms -- Abbreviations -- Preface -- References -- Part I. Introduction and background -- Chapter 1. Introduction: Swedish FrameNet++ -- 1. The Swedish FrameNet++ -- 2. Rationale and aims of SweFN++ -- 2.1 From corpus-based lexicography to language technology R&amp -- D -- 2.2 Extending the shelf life of lexical resources -- 2.3 The increasing importance of the lexicon in language technology -- 2.4 A framenet for Swedish -- 2.5 Serendipitous funding and synergies -- 3. The history of Swedish FrameNet++ -- 4. Integration of existing resources -- 5. A new resource: Swedish FrameNet -- 6. Theoretical and methodological considerations -- 6.1 Interlinking of lexical resources -- 6.2 Method matters -- 6.2.1 Zipf to the rescue -- 6.2.2 Towards a general lexical infrastructure: Karp -- 6.3 Linguistic issues -- 6.3.1 Lexicography and (comparative) linguistics -- 6.3.2 Compounds in Swedish FrameNet -- 6.3.3 Multiword expressions -- 6.4 Computational vs. general linguistics -- 7. Similar initiatives -- 7.1 Multilingual wordnets -- 7.2 MTRoget and multilingual FrameNet -- 7.3 Etymological wordnet, IDS/LWT and the concepticon -- 7.4 BabelNet -- Postscript on BabelNet 5 -- 8. Status and future -- 9. This volume -- Funding -- References -- Appendix A. Swedish FrameNet++ - publications -- Chapter 2. Swedish FrameNet -- 1. Introduction -- 1.1 Berkeley FrameNet -- 1.2 International framenets -- 2. Framenet development methodologies -- 2.1 The extension approach -- 2.2 Merging approach -- 2.3 Modified and new frames -- 3. Language resources and tools for building SweFN -- 4. The SweFN database -- 4.1 Database fields -- 4.2 Annotation and encoding of the data -- 4.3 Consistency checks and evaluation -- 5. Concluding remarks -- Acknowledgements -- Funding.

References -- Part II. Harmonization and integration -- Chapter 3. Swedish FrameNet++ - lexical samsara -- 1. Introduction -- 2. Saldo: The heart of Swedish FrameNet++ -- 2.1 Saldo in a nutshell -- 2.2 The origin of Saldo -- 2.3 The semantic structure of Saldo -- 2.4 Morphological information in Saldo -- 3. Persistent identifiers: The glue of Swedish FrameNet++ -- 4. Branching out: Lexical semantics galore -- 4.1 The Swesaurus component of Swedish FrameNet++ -- 4.2 Towards a thesaurus component of Swedish FrameNet++ -- 5. Looking forward: New directions up ahead -- Acknowledgements -- Funding -- References -- Chapter 4. A lexical resource for computational historical linguistics -- 1. Introduction -- 2. A brief overview of Swedish language stages -- 3. Diachronical lexical resources -- 3.1 Adding diachronical lexicons to SweFN++ -- 3.2 A lexical resource for Late Modern Swedish -- 3.3 A lexical resource for Early Modern Swedish -- 3.4 A lexical resource for Old Swedish -- 4. Diapivot -- 4.1 Methods of automatically linking lexical resources -- 4.2 An application: Studying lexical change and grammaticalization -- 5. Spelling variation and linking texts to lexicons -- 5.1 A noisy channel approach to lemmatization -- 5.2 Training a model on dictionary data -- 5.3 Evaluation -- 5.4 An application: FSvReader -- 6. Conclusions -- Funding -- References -- Appendix A. Definition of the lemmatization model -- Chapter 5. A multilingual net of lexical resources -- 1. Introduction -- 2. Synonyms and their translations -- 3. Lexical resources and their inter-lingual relations -- 3.1 Danish -- 3.2 Estonian -- 3.3 Finnish -- 3.4 Norwegian -- 3.5 Swedish -- 3.6 Multi-lingual visualization using WordTies -- 4. Conclusion -- Acknowledgements -- References -- Chapter 6. Swedish FrameNet++ and comparative linguistics -- 1. The multilingual aspects of Swedish FrameNet++.

2. Core vocabularies for comparative linguistic studies -- 2.1 Basic vocabularies in linguistics -- 2.2 The composition and size of core vocabularies -- 3. Two lexical databases for investigation of South Asian linguistic diversity and unity -- 3.1 Linguistic diversity in South Asia -- 3.2 Grierson's comparative vocabulary in Swedish FrameNet++ -- 3.3 The Intercontinental Dictionary Series as a comparative linguistic research tool -- 3.3 The Intercontinental Dictionary Series as a comparative linguistic research tool -- 4. Conclusion and future prospects -- Acknowledgements -- Funding -- References -- Part III. Method development -- Chapter 7. NLP for resource building -- 1. Introduction -- 1.1 Frame semantics and frame-semantic lexicons -- 2. Computational representation of the meaning of words -- 2.1 The semantic network Saldo -- 2.2 Semantic representations induced from corpora -- 3. From word meaning to frame meaning -- 3.1 Methods based on distance and similarity measures -- 3.2 Classification-based methods -- 4. Quantitative evaluation -- 4.1 Evaluation metrics -- 4.2 Which way is the best to make use of the Saldo lexicon? -- 4.3 Which corpus-based semantic representations are most effective? -- 4.4 Combining lexicon-based and corpus-based classifiers -- 4.5 For which frames are our methods successful? -- 4.6 Use by lexicographers -- 5. Conclusion -- Acknowledgements -- Funding -- References -- Chapter 8. Differing design decisions - comparing Swedish FrameNet to FrameNet -- 1. Introduction -- 2. Differences -- 3. Linking to a dictionary -- 4. New frames for additional concepts -- 5. Polysemy -- 5.1 Hyponymy relations -- 5.2 Regular polysemy and Guest_LUs -- 5.3 Diverse meaning potentials -- 5.4 Frame relations and potential meanings -- 5.5 Complex relations -- 5.6 Polysemy and Swedish FrameNet: Summing up -- 6. Compounds.

6.1 Non-compositional compounds -- 6.2 Compositional compounds -- 6.3 Partially transparent compounds -- 6.4 The constituent-affix cline -- 7. Lexical incorporation of frame element -- 8. Socio-cultural differences -- 9. Conclusions -- Acknowledgements -- Funding -- References -- Chapter 9. Multiword expressions - a tough typological nut for Swedish FrameNet++ -- 1. Background -- 2. Multiword expressions in Swedish FrameNet++ -- 3. MWEs from a typological perspective: A first cut -- 3.1 The "words" of MWEs -- 3.2 The "lexemes" of MWEs -- 3.3 How frequent are multiword expressions in language? -- 3.4 What kinds of MWEs are there? -- 3.5 Where do we find cross-linguistic MWE data? -- 4. Taking stock: Towards a typology of MWEs? -- Acknowledgements -- Funding -- References -- Part IV. Natural language processing applications -- Chapter 10. Semantic role labeling -- 1. Introduction -- 2. The Swedish FrameNet -- 3. Semantic role labeling with SweFN -- 3.1 Segmentation and labeling classifiers -- 4. Experiments -- 4.1 Experimental data and preprocessing -- 4.2 Cross-validation over sentences -- 4.3 Cross-frame role label generalization -- 4.4 Analysis of features -- 4.5 Cross-validation over frames -- 4.6 Increasing classifier robustness by adding cluster features -- 4.7 The effect of syntactic parser choice -- 4.8 Evaluation in the medical domain -- 4.9 Summary of results for the baseline systems -- 5. Using the FrameNet relational structure to improve the semantic role labeler -- 5.1 A classifier using non-atomic semantic role labels -- 5.2 Generalization methods -- 6. Experiments in cross-frame generalization -- 7. Conclusion -- Acknowledgements -- References -- Chapter 11. Computational representation of FrameNet for multilingual natural language generation -- 1. Introduction -- 2. Comparison of selected framenets -- 2.1 Berkeley FrameNet.

2.2 Swedish FrameNet -- 2.3 Summary of the comparison -- 3. Computational framenets in Grammatical Framework -- 3.1 Grammatical Framework -- 3.2 FrameNet grammar library in GF -- 3.3 Status of the FrameNet grammar library -- 4. FrameNet-based multilingual NLG -- 4.1 Accurate generation of tourist phrases -- 4.2 Coherent text generation of museum objects -- 5. Final words -- Funding -- References -- Appendix A. Brief introduction to the GF Resource Grammar Library -- Chapter 12. Language learning and teaching with Swedish FrameNet++: Two examples -- 1. Introduction -- 1.1 Language technology and language pedagogy -- 2. Using resources within SweFN++ for learning and teaching language proficiency and grammatical analysis -- 2.1 The Swedish constructicon as a pedagogical resource -- 2.2 Exploring the usefulness of SweCcn and construction grammar for the teaching of Swedish as a second language -- 2.3 Pattern finding -- 2.4 Type case -- 2.5 Applying construction-based L2-teaching in the classroom - two small-scale studies -- 2.6 SweFN for learning linguistic analysis - semantic roles in Lärka -- 3. Developing the language pedagogical potential within SweFN++ -- 4. Concluding remarks -- Acknowledgements -- Funding -- References -- Index.