---
title: "BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for"
description: "The paper presents the BuzzSaw entry to SemEval-2017 Task 7. 10 references, cited by 3 papers here."
authors: ["Dieke Oele", "Kilian Evang"]
year: 2017
venue: "Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)"
source: "https://aclanthology.org/S17-2076/"
doi: "10.18653/v1/s17-2076"
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/oele-2017-buzzsaw-semeval-2017-task-7/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/oele-2017-buzzsaw-semeval-2017-task-7/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2017](https://md.research.elamur.ai/free/en/papers/2017/page/5/) / BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for Interpreting and Locating Homographic English Puns with Sense Embeddings

# BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for Interpreting and Locating Homographic English Puns with Sense Embeddings

**Authors:** Dieke Oele and [Kilian Evang](https://md.research.elamur.ai/free/en/authors/kilian-evang/) · **Year:** 2017 · **Venue:** [Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)](https://md.research.elamur.ai/free/en/venues/semeval/) ([SemEval 2017](https://md.research.elamur.ai/free/en/venues/semeval/2017/)) · **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/S17-2076/) · [PDF](https://aclanthology.org/S17-2076.pdf) · [DOI 10.18653/v1/s17-2076](https://doi.org/10.18653/v1/s17-2076) · [OpenAlex](https://openalex.org/W2752334231)

**Citations:** 7 citations (OpenAlex, 2026-10-14) · [3](https://md.research.elamur.ai/free/en/papers/oele-2017-buzzsaw-semeval-2017-task-7/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper presents the BuzzSaw entry to SemEval-2017 Task 7. It covers two subtasks for homographic English puns: interpretation (naming the two WordNet senses a pun evokes) and location (finding which word is the pun). The core is Lesk++, a knowledge-based WSD method. It scores each candidate sense by comparing a context vector with the sense's expanded gloss vector and with its lexeme embedding. The embeddings come from AutoExtend on top of word2vec Google News vectors. The authors observe that a joke usually holds cues for each sense in separate stretches of text. So instead of taking the top two senses for the whole sentence, they cut the context into two parts and disambiguate the pun once against each part. For location, they choose the polysemous word whose two assigned senses lie farthest apart in embedding space. Splitting the context beats the unsplit top-two approach, and the system beats random and MFS baselines on interpretation. For location, it beats random choice but loses clearly to a trivial "last content word" heuristic.

## Key points
- Lesk++ scores a sense as the cosine of the gloss vector with the context plus the cosine of the lexeme vector with the context.
  - Glosses are enlarged with the glosses of related WordNet synsets, leaving out antonyms.
  - Words are disambiguated from fewest senses to most, and words already resolved contribute their sense vectors to the context.
- Resources: Princeton WordNet 3.1, AutoExtend sense/lexeme embeddings sharing the space of 300-dimensional word2vec Google News vectors, and the Stanford POS tagger. Development used 91 puns from "Pun of the Day", annotated by both authors with adjudication.
- Interpretation precision:
  - random 7.24
  - MFS 11.21
  - Lesk++ without splitting 15.45
  - split in half 16.39 (best)
  - split by punctuation 15.53
  - max-dissimilarity split 15.53 (the submitted run)
  - Coverage was about 98%.
- Location precision:
  - random content word 13.20
  - Lesk++ max-dissimilarity 27.69
  - last content word 52.96, because puns usually sit at the end of the joke.
- The quality of interpretation output had little effect on location (about 27.7–28.4 for all Lesk++ variants, versus 17–18 for random/MFS senses). Puns whose senses differ in part of speech, phrasal verbs and MWEs are not handled.

## Interesting ideas
- When a word should carry two readings, splitting the text into two local contexts and disambiguating against each gives sharper evidence than taking the runner-up sense from a single global context.
- A pun can be located by searching for the word whose two best senses are most semantically distant, turning interpretation output directly into a location signal.
- Disambiguating the least ambiguous words first and feeding their chosen senses back into the context is a cheap, explainable bootstrapping order.
- A naive positional heuristic (last content word) crushed the semantic method, which shows that simple structural baselines must be part of any evaluation.
- Several context-splitting strategies (half, punctuation, max-dissimilarity) were compared, and the simplest one won, which argues for ablating clever segmentation against blunt segmentation.

## For Mova

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

- **I1** (frames): Irony and pun detection in prag could run sense or frame selection separately on the setup and the punchline, and flag a double reading when the two winners differ.
- **I2** (world): When building the world of a joke or riddle, an entity or verb with two strongly divergent WordNet senses could be marked as a deliberate ambiguity point instead of being forced to one reading.
- **I3** (world): Sense assignment over the WordNet ontology could follow this easiest-first order, logging which resolved neighbours justified each later choice.
- **I4** (learn): Evaluation suites should always include trivial positional or frequency baselines as negative controls, so that dataset artefacts are not mistaken for real understanding.
- **I5** (answer): When readers pick the passage span relevant to a question, they should be benchmarked against plain halving or sentence-boundary splits before more elaborate segmenters are adopted.

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

## Institutions

- [University of Groningen](https://md.research.elamur.ai/free/en/universities/university-of-groningen/) — University, Groningen, Netherlands · 74 works · 5,142 citations

📚 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 (2): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works · [📖 Lexical semantics & word embeddings](https://md.research.elamur.ai/free/en/topics/lexical-semantics/) · 65 works

## Cite

Cite as: Dieke Oele and Kilian Evang. 2017. BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for Interpreting and Locating Homographic English Puns with Sense Embeddings. Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017).

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

```bibtex
@inproceedings{oele2017buzzsaw,
  title = {{BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for Interpreting and Locating Homographic English Puns with Sense Embeddings}},
  author = {Dieke Oele and Kilian Evang},
  booktitle = {Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)},
  year = {2017},
  doi = {10.18653/v1/s17-2076},
  url = {https://doi.org/10.18653/v1/s17-2076}
}
```

- [← previous Modeling Dialogue Acts with Content Word Filtering and Speaker Preferences](https://md.research.elamur.ai/free/en/papers/jo-2017-modeling-dialogue-acts-content-word/)
- [next → Developing a concept-level knowledge base for sentiment analysis in Singlish](https://md.research.elamur.ai/free/en/papers/bajpai-2017-developing-concept-level-knowledge-base/)
- [↑ back to papers of 2017](https://md.research.elamur.ai/free/en/papers/2017/page/5/)

## References

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

1. Kristina Toutanova, Dan Klein, Christopher D. Manning, and Yoram Singer. 2003. [Feature-rich part-of-speech tagging with a cyclic dependency network](https://md.research.elamur.ai/free/en/papers/toutanova-2003-feature-rich-part-speech-tagging/). In Proceedings of the 2003 Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics. http://aclweb.org/anthology/N03-1033. [↗](https://aclanthology.org/N03-1033/)
2. Satanjeev Banerjee and Ted Pedersen. 2002. An adapted Lesk algorithm for word sense disambiguation using WordNet. In Proceedings of the Third International Conference on Computational Linguistics and Intelligent Text Processing. Springer-Verlag, London, UK, UK, CICLing ’02, pages 136–145. [http://dl.acm.org/citation.cfm?id=647344.724142](http://dl.acm.org/citation.cfm?id=647344.724142). [↗](http://dl.acm.org/citation.cfm?id=647344.724142)
3. Xinxiong Chen, Zhiyuan Liu, and Maosong Sun. 2014. [A Unified Model for Word Sense Representation and Disambiguation](https://md.research.elamur.ai/free/en/papers/chen-2014-unified-model-word-sense-representation/). In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, EMNLP 2014, October 25-29, 2014, Doha, Qatar, A meeting of SIGDAT, a Special Interest Group of the ACL. pages 1025–1035. http://aclweb.org/anthology/D/D14/D14-1110.pdf. [↗](https://aclanthology.org/D14-1110/)
4. Christiane Fellbaum, editor. 1998. WordNet An Electronic Lexical Database. The MIT Press, Cambridge, MA ; London. [find](https://scholar.google.com/scholar?q=Christiane+Fellbaum%2C+editor%2E+1998%2E+WordNet+An+Electronic+Lexical+Database%2E+The+MIT+Press%2C+Cambridge%2C+MA+%3B+London%2E)
5. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. [Efficient estimation of word representations in vector space](https://md.research.elamur.ai/free/en/papers/mikolov-2013-efficient-estimation-word-representations-vector/). CoRR abs/1301.3781. http://dblp.uni-trier.de/db/journals/corr/corr1301.html. [↗](https://arxiv.org/abs/1301.3781)
6. Tristan Miller and Iryna Gurevych. 2015. [Automatic disambiguation of English puns](https://md.research.elamur.ai/free/en/papers/miller-2015-automatic-disambiguation-english-puns/). In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Association for Computational Linguistics, Beijing, China, pages 719–729. http://www.aclweb.org/anthology/P15-1070. [↗](https://aclanthology.org/P15-1070/)
7. Tristan Miller, Christian F. Hempelmann, and Irina Gurevych. 2017. [Semeval-2017 task 7: Detection and interpretation of English puns](https://md.research.elamur.ai/free/en/papers/miller-2017-semeval-2017-task-7-detection/). In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017).
8. Tristan Miller and Mladen Turković. 2016. Towards the automatic detection and identification of English puns. European Journal of Humour Research 4. [find](https://scholar.google.com/scholar?q=Tristan+Miller+and+Mladen+Turkovic%CC%81%2E+2016%2E+Towards+the+automatic+detection+and+identification+of+English+puns%2E+European+Journal+of+Humour+Research+4%2E)
9. Dieke Oele and Gertjan van Noord. 2017. Distributional Lesk: Effective multilingual knowledge-based word sense disambiguation. Submitted for publication. [find](https://scholar.google.com/scholar?q=Dieke+Oele+and+Gertjan+van+Noord%2E+2017%2E+Distributional+Lesk%3A+Effective+multilingual+knowledge%2Dbased+word+sense+disambiguation%2E+Submitted+for+publication%2E)
10. Sascha Rothe and Hinrich Schütze. 2015. [Autoextend: Extending word embeddings to embeddings for synsets and lexemes](https://md.research.elamur.ai/free/en/papers/rothe-2015-autoextend-extending-word-embeddings-embeddings/). In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Association for Computational Linguistics, pages 1793–1803. https://doi.org/10.3115/v1/P15-1173. [↗](https://doi.org/10.3115/v1/p15-1173)

*Updated 2 November 2026*