Adaptation of Morphosyntactic Taggers

Scherrer, Yves

doi:10.1017/9781108565080.011

Cited by 5 publications

(17 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One possible pick that satisfies all the requirements is the Stanza tagger, developed at Stanford for Universal Dependencies tagging (Qi et al, 2018) . It was recently modified to use bidirectional character-level LSTM by default, and specifically adjusted to the aims of part-of-speech tagging, the starting point for lowresource NLP (Scherrer, 2021). This fork is used in this paper.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Tenth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2023)

2023

View full text Add to dashboard Cite

Hate speech detection in online platforms has been widely studied in the past. Most of these works were conducted in English and a few rich-resource languages. Recent approaches tailored for low-resource languages have explored the interests of zero-shot cross-lingual transfer learning models in resource-scarce scenarios. However, languages variations between geolects such as American English and British English, Latin-American Spanish, and European Spanish is still a problem for NLP models that often relies on (latent) lexical information for their classification tasks. More importantly, the cultural aspect, crucial for hate speech detection, is often overlooked.In this work, we present the results of a thorough analysis of hate speech detection models performance on different variants of Spanish, including a new hate speech toward immigrants Twitter data set we built to cover these variants. Using mBERT and Beto, a monolingual Spanish Bert-based language model, as the basis of our transfer learning architecture, our results indicate that hate speech detection models for a given Spanish variant are affected when different variations of such language are not considered. Hate speech expressions could vary from region to region where the same language is spoken. * Work conducting during an internship at Inria Paris. 1 Please be aware that this paper contains some examples of offensive slurs that may be considered upsetting. ReferencesResham Ahluwalia, Himani Soni, Edward Callow, Anderson Nascimento, and Martine De Cock. 2018. Detecting hate speech against women in english tweets. EVALITA Evaluation of NLP and Speech Tools for Italian, 12:194.

show abstract

Section: Methodsmentioning

confidence: 99%

“…Now a lot of attention is paid to the selection of data to train, evaluate, and test a tagger on (Muradoglu and Hulden, 2022). The old models (Qi et al, 2018) are adjusted (Scherrer, 2021) to meet the new requirements of efficient training on low-resourced closely-related lects.…”

Section: Related Workmentioning

confidence: 99%

Tenth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2023)

2023

View full text Add to dashboard Cite

show abstract

“…Part-ofspeech tagging is mostly dominated by universal methods, based on recurrent neural networks (Qi et al, 2018) (Qi et al, 2020). Yet the tasks conducted on different language varieties demand agile models that can both be tuned for the needs of a specific tagset and work in the context of lowresourced and sparse data (Scherrer, 2021). Hidden Markov Model (HMM)-based taggers present this opportunity (Schmid, 1994(Schmid, , 1995Özçelik et al, 2019;Lyashevskaya and Afanasev, 2021).…”

Section: Related Workmentioning

confidence: 99%

Multi-lect automatic detection of Swadesh list items from raw corpus data in East Slavic languages

Afanasev

2023

Proceedings of the 4th Workshop on Computational Approaches to Historical Language Change

View full text Add to dashboard Cite

The article introduces a novel task of multilect automatic detection of Swadesh list items from raw corpora. The task aids the early stage of historical linguistics study by helping the researcher compile word lists for further analysis.In this paper, I test multi-lect automatic detection on the East Slavic lects' data. The training data consists of Ukrainian, Belarusian, and Russian material. I introduce a new dataset for the Ukrainian language. I implement data augmentation techniques to give automatic tools a better understanding of the searched value. The test data consists of the Old East Slavic texts.I train HMM, CRF, and mBERT models, then test and evaluate them by harmonic F1 score. The baseline is a Random Forest classifier. I introduce two different subtasks: the search for new Swadesh list items, and the search for the known Swadesh list items in new lects of the well-established group. The first subtask, given the simultaneously diverse and vague nature of the Swadesh list, currently presents an almost unbeatable challenge for machine learning methods. The second subtask, on the other hand, is easier, and the mBERT model achieves a 0.57 F1 score. This is an impressive result, given how hard it is to formalise the token belonging to a very specific and thematically diverse set of concepts.

show abstract

“…Part-ofspeech tagging is mostly dominated by universal methods, based on recurrent neural networks (Qi et al, 2018) ). Yet the tasks conducted on different language varieties demand agile models that can both be tuned for the needs of a specific tagset and work in the context of lowresourced and sparse data (Scherrer, 2021). Hidden Markov Model (HMM)-based taggers present this opportunity (Schmid, 1994(Schmid, , 1995Özçelik et al, 2019;Lyashevskaya and Afanasev, 2021).…”

Section: Related Workmentioning

confidence: 99%

“…This requires using special metrics for imbalanced data (Dudy and Bedrick, 2020). The harmonic F1 score, traditionally used for such cases (Chinchor, 1992), still finds its application in the analysis of NLP tasks (Scherrer, 2021).…”

Section: Related Workmentioning

confidence: 99%

Proceedings of the 4th Workshop on Computational Approaches to Historical Language Change

2023

View full text Add to dashboard Cite

Welcome to the 4th International Workshop on Computational Approaches to Historical Language Change (LChange'23) co-located with EMNLP 2023. LChange is held on December 6th, 2023, as a hybrid event with participation possible both virtually and on-site in Singapore.Characterizing the time-varying nature of language will have broad implications and applications in multiple fields including linguistics, artificial intelligence, digital humanities, computational cognitive and social sciences. In this workshop, we bring together the world's pioneers and experts in computational approaches to historical language change with a focus on digital text corpora. In doing so, this workshop carries out the triple goals of disseminating state-of-the-art research on diachronic modeling of language change, fostering cross-disciplinary collaborations, and exploring the fundamental theoretical and methodological challenges in this growing niche of computational linguistic research.In response to the call, we received 28 submissions. Each of them was carefully evaluated by at least two members of the Program Committee, whom we believed to be most appropriate for each paper. Based on the reviewers' feedback we accepted 17 full and short papers as oral or poster presentations. We had two distinguished keynote presentations: the first by Gemma Boleda (Research Professor in the Department of Translation and Language Sciences of the Universitat Pompeu Fabra, Spain) who presented a talk entitled "What does semantic change have to do with Hello Kitty? Referring as the source of change", and the second by Mario Giulianelli (a postdoctoral fellow at ETH Zurich) with the talk "Neural language models for word usage representation and analysis". Finally, we invited five EMNLP'23 Findings papers to be presented as posters, which are not included in the workshop proceedings.To further support the community, we offered five student scholarships to cover registration fees. We also offered mentoring for four young researchers on their research topic in the field of language change, either during the workshop or virtually.We hope that you will find the workshop papers insightful and inspiring. We would like to thank the keynote speakers for their stimulating talks, the authors of all papers for their interesting contributions, and the members of the Program Committee for their insightful reviews. Our special thanks go to the emergency reviewers who stepped in to provide their expertise. We also express our gratitude to the EMNLP 2023 workshop chairs for their kind assistance during the organization process. Finally, our thanks go to our gold sponsor iguanodon.ai, as well as the research project "Towards Computational Lexical Semantic Change Detection" (Swedish Research Council, contract 2018-01184) and the research program "Change is Key!" (Riksbankens Jubileumsfond, contract M21-0021).

show abstract

Adaptation of Morphosyntactic Taggers

Cited by 5 publications

References 46 publications

Tenth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2023)

Tenth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2023)

Multi-lect automatic detection of Swadesh list items from raw corpus data in East Slavic languages

Proceedings of the 4th Workshop on Computational Approaches to Historical Language Change

Contact Info

Product

Resources

About