---
title: "JCT at SemEval-2022 Task 4-A: Patronism Detection in Posts"
description: "The paper describes the JCT team's entry for SemEval-2022 Task 4, subtask A. The subtask is a binary… 0.3536 recall. 27 references."
authors: ["Yaakov HaCohen-Kerner", "Ilan Meyrowitsch", "Matan Fchima"]
year: 2022
venue: "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)"
source: "https://aclanthology.org/2022.semeval-1.72/"
doi: "10.18653/v1/2022.semeval-1.72"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/kerner-2022-jct-semeval-2022-task-4/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/kerner-2022-jct-semeval-2022-task-4/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2022](https://md.research.elamur.ai/free/en/papers/2022/page/10/) / JCT at SemEval-2022 Task 4-A: Patronism Detection in Posts Written in English using Preprocessing Methods and various Machine Leaerning Methods

# JCT at SemEval-2022 Task 4-A: Patronism Detection in Posts Written in English using Preprocessing Methods and various Machine Leaerning Methods

**Authors:** [Yaakov HaCohen-Kerner](https://md.research.elamur.ai/free/en/authors/yaakov-hacohen-kerner/), [Ilan Meyrowitsch](https://md.research.elamur.ai/free/en/authors/ilan-meyrowitsch/) and [Matan Fchima](https://md.research.elamur.ai/free/en/authors/matan-fchima/) · **Year:** 2022 · **Venue:** [Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)](https://md.research.elamur.ai/free/en/venues/semeval/) ([SemEval 2022](https://md.research.elamur.ai/free/en/venues/semeval/2022/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2022.semeval-1.72/) · [PDF](https://aclanthology.org/2022.semeval-1.72.pdf) · [DOI 10.18653/v1/2022.semeval-1.72](https://doi.org/10.18653/v1/2022.semeval-1.72) · [OpenAlex](https://openalex.org/W4287889453)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper describes the JCT team's entry for SemEval-2022 Task 4, subtask A. The subtask is a binary decision: does a paragraph from English news text contain patronizing or condescending language (PCL)? The data is the Don't Patronize Me! corpus. It is strongly skewed, with only about one example in ten labelled PCL.

The authors compared 11 supervised learners and 9 text-cleaning steps:
- **Classical:** RF, KNN, SVC, XGBoost, LR, Decision Tree, Naive Bayes.
- **Transformer:** Bert, DistilBert, Roberta, Albert.

To reduce the skew, they added extra positive examples. These were fragments cut from the span-level PCL annotations used in the multi-label subtask. Their strongest system was a fine-tuned BertForSequenceClassification model. It reached a PCL-class F1 of 0.77 on their internal split but only 0.377 on the official test set, ranking 62nd.

The work matters mainly as a cautionary baseline. It shows that rebalancing and cleaning text help locally. It also shows that a large gap between internal and official scores exposes distribution shift for a subtle, context-dependent phenomenon like condescension.

## Key points
- **Training data:** 10,469 paragraphs, of which 993 are PCL (about 9.5%). The original ratio is roughly 91:9.
- **Rebalancing:** adding partial positive fragments from the Task 4 category annotations moved the ratio to 77:23. A fully balanced 50:50 set built by duplicating positives gave worse results than this unequal ratio.
- **Preprocessing:** nine cleaning steps were used:
  - removing newlines, HTML tags, links, extra whitespace, accented characters and special characters;
  - lower-casing;
  - reducing repeated characters and punctuation;
  - expanding contractions.
- **Best model ("Matan_bert"):**
  - Settings: learning rate 2e-7, epsilon 1e-8, 20 epochs, batch size 16. The paper describes the model as having "768 layers", which is most likely the hidden size.
  - Training stopped at 20 epochs because the gap between training and validation loss began to widen.
  - Official test results: precision 0.3536, recall 0.4038, F1 0.377.
- **Second submission:** a Decision Tree (entropy criterion, 70:30 split) scored 0.7 F1 internally; its test score was not published. Code is available at github.com/meyrow/pcl-detection-task4-semeval2022.

## Interesting ideas
- Grow the rare class from finer-grained annotations: cut the annotated positive spans out of longer texts as new examples, rather than copying whole examples, which overfit more in their tests.
- A moderately rebalanced set (77:23) beat a perfectly balanced duplicated one (50:50), so the target class ratio deserves to be tuned like a hyperparameter.
- The drop from 0.77 to 0.377 F1 shows that an internal random split can badly overestimate performance on a shifted official test set.
- Condescension is characterised by three traits: no overtly negative content (often insincere praise), an assumed status gap between writer and subject, and a listener who rejects that gap.
- Normalisation steps such as contraction expansion and collapsing repeated characters are applied as a separate, systematically compared pipeline stage.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): when silver data is thin for a rare construction (e.g., a speech act or ironic frame), annotated sub-spans from richer corpora could seed extra training items for the witness parsers.
- **I2** (learn): when training classifiers for rare frame or role labels, try several minority ratios against a held-out set instead of forcing full balance.
- **I3** (learn): evaluation should include out-of-domain held-out texts and negative controls, so that a score gain from memorised topic or style is caught early.
- **I4** (frames): these traits could become explicit, explainable checks in the pragmatics layer, e.g., positive evaluation combined with an inferiority implicature about a group.
- **I5** (tokens): the en tokenizer could adopt such normalisations as reversible, offset-preserving rewrites, so that noisy web text is cleaned without losing the path back to the original string.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works

## Cite

Cite as: Yaakov HaCohen-Kerner, Ilan Meyrowitsch and Matan Fchima. 2022. JCT at SemEval-2022 Task 4-A: Patronism Detection in Posts Written in English using Preprocessing Methods and various Machine Leaerning Methods. Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{hacohenkerner2022jct,
  title = {{JCT at SemEval-2022 Task 4-A: Patronism Detection in Posts Written in English using Preprocessing Methods and various Machine Leaerning Methods}},
  author = {Yaakov HaCohen-Kerner, Ilan Meyrowitsch and Matan Fchima},
  booktitle = {Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)},
  year = {2022},
  doi = {10.18653/v1/2022.semeval-1.72},
  url = {https://doi.org/10.18653/v1/2022.semeval-1.72}
}
```

- [← previous ISD at SemEval-2022 Task 6: Sarcasm Detection Using Lightweight Models](https://md.research.elamur.ai/free/en/papers/huang-2022-isd-semeval-2022-task-6/)
- [next → Lexical and functional adpositions: the view from <em>of</em> in Old and present-day English](https://md.research.elamur.ai/free/en/papers/taylor-2022-lexical-functional-adpositions-view-em/)
- [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/page/10/)

## References

27 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Nitesh V. Chawla, Kevin W Bowyer, Lawrence O. Hall, and W. Philip Kegelmeyer. 2002. [SMOTE: synthetic minority over-sampling technique](https://md.research.elamur.ai/free/en/papers/chawla-2002-smote-synthetic-minority-over-sampling/). Journal of artificial intelligence research, 16, 321-357.
2. Valerio Basile, Cristina Bosco, Elisabetta Fersini, Debora Nozza, Viviana Patti, Francisco Manuel Rangel Pardo, Paolo Rosso, and Manuela Sanguinetti. 2019. [SemEval-2019 task 5: Multilingual detection of hate speech against immigrants and women in twitter](https://md.research.elamur.ai/free/en/papers/basile-2019-semeval-2019-task-5-multilingual/). In Proceedings of SemEval.
3. Niall J. Conroy, Victoria L. Rubin, and Yimin Chen. 2015. Automatic deception detection: Methods for finding fake news. Proceedings of the Association for Information Science and Technology, 52(1):1–4. [find](https://scholar.google.com/scholar?q=Niall+J%2E+Conroy%2C+Victoria+L%2E+Rubin%2C+and+Yimin+Chen%2E+2015%2E+Automatic+deception+detection%3A+Methods+for+finding+fake+news%2E+Proceedings+of+the+Association+for+Information+Science+and+Technology%2C+52%281%29%3A1%E2%80%934)
4. Jason Brownlee. 2020. Imbalanced classification with python: Better metrics, balance skewed classes, cost-sensitive learning. Machine Learning Mastery. [find](https://scholar.google.com/scholar?q=Jason+Brownlee%2E+2020%2E+Imbalanced+classification+with+python%3A+Better+metrics%2C+balance+skewed+classes%2C+cost%2Dsensitive+learning%2E+Machine+Learning+Mastery%2E)
5. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [BERT: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics. 19th Industrial Conference on Data Mining, (ICDM 2019), New York.
6. Yaakov HaCohen-Kerner, Daniel Miller, and Yair Yigal. 2020. The influence of preprocessing on text classification using a bag-of-words representation, PloS one, vol. 15, p. e0232525. [find](https://scholar.google.com/scholar?q=Yaakov+HaCohen%2DKerner%2C+Daniel+Miller%2C+and+Yair+Yigal%2E+2020%2E+The+influence+of+preprocessing+on+text+classification+using+a+bag%2Dof%2Dwords+representation%2C+PloS+one%2C+vol%2E+15%2C+p%2E+e0232525%2E)
7. Bartosz Krawczyk. 2016. Learning from imbalanced data: open challenges and future directions. Progress in Artificial Intelligence, 5(4), 221-232. [find](https://scholar.google.com/scholar?q=Bartosz+Krawczyk%2E+2016%2E+Learning+from+imbalanced+data%3A+open+challenges+and+future+directions%2E+Progress+in+Artificial+Intelligence%2C+5%284%29%2C+221%2D232%2E)
8. Haibo He and Yunqian Ma (Eds.). 2013. Imbalanced learning: foundations, algorithms, and applications. [find](https://scholar.google.com/scholar?q=Haibo+He+and+Yunqian+Ma+%28Eds%2E%29%2E+2013%2E+Imbalanced+learning%3A+foundations%2C+algorithms%2C+and+applications%2E)
9. Thomas Huckin. 2002. Critical discourse analysis and the discourse of condescension. Discourse studies in composition, 155, 176. [find](https://scholar.google.com/scholar?q=Thomas+Huckin%2E+2002%2E+Critical+discourse+analysis+and+the+discourse+of+condescension%2E+Discourse+studies+in+composition%2C+155%2C+176%2E)
10. Ven-hwei Lo, and Ran Wei. 2006. Perceptual differences in assessing the harm of patronizing adult entertainment clubs. International Journal of Public Opinion Research 18.4: 475-487. [find](https://scholar.google.com/scholar?q=Ven%2Dhwei+Lo%2C+and+Ran+Wei%2E+2006%2E+Perceptual+differences+in+assessing+the+harm+of+patronizing+adult+entertainment+clubs%2E+International+Journal+of+Public+Opinion+Research+18%2E4%3A+475%2D487%2E)
11. Aditya Joshi, Pushpak Bhattacharyya, and Mark J. Carman. 2017. [Automatic sarcasm detection: A survey](https://md.research.elamur.ai/free/en/papers/joshi-2016-automatic-sarcasm-detection-survey/). ACM Computing Surveys (CSUR), 50(5), 1-22.
12. Yang Liu, Elizabeth Shriberg, Andreas Stolcke, and Mary Harper. 2004. Using machine learning to cope with imbalanced classes in natural speech: Evidence from sentence boundary and disfluency detection. In Eighth International Conference on Spoken Language Processing. [find](https://scholar.google.com/scholar?q=Yang+Liu%2C+Elizabeth+Shriberg%2C+Andreas+Stolcke%2C+and+Mary+Harper%2E+2004%2E+Using+machine+learning+to+cope+with+imbalanced+classes+in+natural+speech%3A+Evidence+from+sentence+boundary+and+disfluency+detection)
13. Yaakov HaCohen-Kerner, Ariel Kass, and Ariel Peretz. 2008. Combined one sense disambiguation of abbreviations. In Proceedings of ACL-08: HLT, Short Papers, Association for Computational Linguistics, pages 61-64, Columbus, Ohio, Association for Computational Linguistics. URL: https://aclanthology.org/P08-2. [↗](https://aclanthology.org/P08-2)
14. Shirley D. McCune and Martha Matthews. 1978. Implementing Title IX and attaining sex equity: a workshop package for elementary-secondary educators: the community's role: outline and participants' materials for application sessions for community group members. Department of Health, Education, and Welfare,[Education Division], Office of Education. [find](https://scholar.google.com/scholar?q=Shirley+D%2E+McCune+and+Martha+Matthews%2E+1978%2E+Implementing+Title+IX+and+attaining+sex+equity%3A+a+workshop+package+for+elementary%2Dsecondary+educators%3A+the+community%27s+role%3A+outline+and+participants%27+mate)
15. Yaakov HaCohen-Kerner, Ittay Stern, David Korkus, and Erick Fredj. 2007. Automatic machine learning of keyphrase extraction from short HTML documents written in Hebrew. Cybernetics and Systems: An International Journal, 38(1), 1-21. [find](https://scholar.google.com/scholar?q=Yaakov+HaCohen%2DKerner%2C+Ittay+Stern%2C+David+Korkus%2C+and+Erick+Fredj%2E+2007%2E+Automatic+machine+learning+of+keyphrase+extraction+from+short+HTML+documents+written+in+Hebrew%2E+Cybernetics+and+Systems%3A+An+Int)
16. Julia Mendelsohn, Yulia Tsvetkov, and Dan Jurafsky. 2020. A framework for the computational linguistic analysis of dehumanization. [find](https://scholar.google.com/scholar?q=Julia+Mendelsohn%2C+Yulia+Tsvetkov%2C+and+Dan+Jurafsky%2E+2020%2E+A+framework+for+the+computational+linguistic+analysis+of+dehumanization%2E)
17. Yaakov HaCohen-Kerner, Dror Mughaz, Hananya Beck, and Elchai Yehudai. 2008. Words as classifiers of documents according to their historical period and the ethnic origin of their authors. Cybernetics and Systems: An International Journal, 39(3), 213-228. [find](https://scholar.google.com/scholar?q=Yaakov+HaCohen%2DKerner%2C+Dror+Mughaz%2C+Hananya+Beck%2C+and+Elchai+Yehudai%2E+2008%2E+Words+as+classifiers+of+documents+according+to+their+historical+period+and+the+ethnic+origin+of+their+authors%2E+Cybernetics+a)
18. Carla Pérez-Almendros, Luis Espinosa-Anke, and Steven Schockaert. 2020. Don't patronize me! An annotated dataset with patronizing and condescending language towards vulnerable communities. In Proceedings of the 28th International Conference on Computational Linguistics. pages 5891-5902. [find](https://scholar.google.com/scholar?q=Carla+P%C3%A9rez%2DAlmendros%2C+Luis+Espinosa%2DAnke%2C+and+Steven+Schockaert%2E+2020%2E+Don%27t+patronize+me%21+An+annotated+dataset+with+patronizing+and+condescending+language+towards+vulnerable+communities%2E+In+Proceedi)
19. Yaakov HaCohen-Kerner, Ariel Kass, and Ariel Peretz. 2010A. HAADS: A Hebrew Aramaic abbreviation disambiguation system. Journal of the American Society for Information Science and Technology, 61(9), 1923-1932. [find](https://scholar.google.com/scholar?q=Yaakov+HaCohen%2DKerner%2C+Ariel+Kass%2C+and+Ariel+Peretz%2E+2010A%2E+HAADS%3A+A+Hebrew+Aramaic+abbreviation+disambiguation+system%2E+Journal+of+the+American+Society+for+Information+Science+and+Technology%2C+61%289%29%2C+1)
20. Carla Pérez-Almendros, Luis Espinosa-Anke, and Steven Schockaert. 2022. SemEval-2022 Task 4: Patronizing and Condescending Language Detection. In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022). Association for Computational Linguistics. Yaakov HaCohen-Kerner, Hananya Beck, Elchai Yehudai, and Dror Mughaz. 2010B. Stylistic feature sets as classifiers of documents according to their historical period and ethnic origin. Applied Artificial Intelligence, 24(9), 847-862. [find](https://scholar.google.com/scholar?q=Carla+P%C3%A9rez%2DAlmendros%2C+Luis+Espinosa%2DAnke%2C+and+Steven+Schockaert%2E+2022%2E+SemEval%2D2022+Task+4%3A+Patronizing+and+Condescending+Language+Detection%2E+In+Proceedings+of+the+16th+International+Workshop+on+Sema)
21. Maarten Sap, Saadia Gabriel, Lianhui Qin, Dan Jurafsky, Noah A Smith, and Yejin Choi. 2019. [Social bias frames: Reasoning about social and power implications of language](https://md.research.elamur.ai/free/en/papers/sap-2020-social-bias-frames-reasoning-about/). arXiv preprint arXiv:1911.03891. [↗](https://arxiv.org/abs/1911.03891)
22. Yaakov HaCohen-Kerner, Ziv Ido, and Ronen Ya’akobov. 2017. Stance classification of tweets using skip char Ngrams. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 266-278). Springer, Cham. [find](https://scholar.google.com/scholar?q=Yaakov+HaCohen%2DKerner%2C+Ziv+Ido%2C+and+Ronen+Ya%E2%80%99akobov%2E+2017%2E+Stance+classification+of+tweets+using+skip+char+Ngrams%2E+In+Joint+European+Conference+on+Machine+Learning+and+Knowledge+Discovery+in+Databases)
23. Zijian Wang and Christopher Potts. 2019. [Talkdown: A corpus for condescension detection in context](https://md.research.elamur.ai/free/en/papers/zampieri-2019-semeval-2019-task-6-identifying/). In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the Yaakov HaCohen-Kerner, Yair Yigal, and Daniel Miller. 2019. The impact of Preprocessing on Classification of Mental Disorders, in Proc. of the 9th International Joint Conference on Natural Language Processing categorizing offensive language in social media (offenseval). arXiv preprint arXiv:1903.08983. [↗](https://arxiv.org/abs/1903.08983)
24. Bee Wah Yap, Khatijahhusna Abd Rani, Hezlin Aryani Abd Rahman, Simon Fong, Zuraida Khairudin, Nik Nairan Abdullah. 2014. An application of oversampling, undersampling, bagging and boosting in handling imbalanced datasets. Proceedings of the first international conference on advanced data and information engineering (DaEng-2013). Springer, Singapore. [find](https://scholar.google.com/scholar?q=Bee+Wah+Yap%2C+Khatijahhusna+Abd+Rani%2C+Hezlin+Aryani+Abd+Rahman%2C+Simon+Fong%2C+Zuraida+Khairudin%2C+Nik+Nairan+Abdullah%2E+2014%2E+An+application+of+oversampling%2C+undersampling%2C+bagging+and+boosting+in+handling)
25. Marcos Zampieri, Preslav Nakov, Sara Rosenthal, Pepa Atanasova, Georgi Karadzhov, Hamdy Mubarak, Leon Derczynski, Zeses Pitenis, and Cagrı Coltekin. 2020. In Proceedings of SemEval Semeval-2020 task 12: Multilingual offensive language identification in social media. [find](https://scholar.google.com/scholar?q=Marcos+Zampieri%2C+Preslav+Nakov%2C+Sara+Rosenthal%2C+Pepa+Atanasova%2C+Georgi+Karadzhov%2C+Hamdy+Mubarak%2C+Leon+Derczynski%2C+Zeses+Pitenis%2C+and+Cagr%C4%B1+Coltekin%2E+2020%2E+In+Proceedings+of+SemEval+Semeval%2D2020+task+1)
26. Tuanfei Zhu, Yaping Lin, and Yonghe Liu. 2017. Synthetic minority oversampling technique for multiclass imbalance problems. Pattern Recognition, 72, 327-340. [find](https://scholar.google.com/scholar?q=Tuanfei+Zhu%2C+Yaping+Lin%2C+and+Yonghe+Liu%2E+2017%2E+Synthetic+minority+oversampling+technique+for+multiclass+imbalance+problems%2E+Pattern+Recognition%2C+72%2C+327%2D340%2E)
27. Marcos Zampieri, Shervin Malmasi, Preslav Nakov, Sara Rosenthal, Noura Farra, and Ritesh Kumar. 2019. Semeval-2019 Task 6: Identifying and [find](https://scholar.google.com/scholar?q=Marcos+Zampieri%2C+Shervin+Malmasi%2C+Preslav+Nakov%2C+Sara+Rosenthal%2C+Noura+Farra%2C+and+Ritesh+Kumar%2E+2019%2E+Semeval%2D2019+Task+6%3A+Identifying+and)

*Updated 2 November 2026*