---
title: "Exploiting Parallel Texts for Word Sense Disambiguation: An Empirical Study — Ng 2003"
description: "Ng, Wang and Chan test whether word-aligned parallel text can supply this training data instead. 14.0 points. 18 references, cited by 67 papers here."
authors: ["Hwee Tou Ng", "Bin Wang", "Yee Seng Chan"]
year: 2003
venue: "Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics"
source: "https://aclanthology.org/P03-1058/"
doi: "10.3115/1075096.1075154"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/ng-2003-exploiting-parallel-texts-word-sense/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/ng-2003-exploiting-parallel-texts-word-sense/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2003](https://md.research.elamur.ai/free/en/papers/2003/) / Exploiting Parallel Texts for Word Sense Disambiguation: An Empirical Study

# Exploiting Parallel Texts for Word Sense Disambiguation: An Empirical Study

**Authors:** [Hwee Tou Ng](https://md.research.elamur.ai/free/en/authors/hwee-tou-ng/), [Bin Wang](https://md.research.elamur.ai/free/en/authors/bin-wang/) and [Yee Seng Chan](https://md.research.elamur.ai/free/en/authors/yee-seng-chan/) · **Year:** 2003 · **Venue:** Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/P03-1058/) · [PDF](https://aclanthology.org/P03-1058.pdf) · [DOI 10.3115/1075096.1075154](https://doi.org/10.3115/1075096.1075154) · [OpenAlex](https://openalex.org/W2145262681)

**Citations:** 142 citations (OpenAlex, 2026-10-14) · [67](https://md.research.elamur.ai/free/en/papers/ng-2003-exploiting-parallel-texts-word-sense/cited-by/) in our library, of the 140,884 papers whose bibliographies we read

## Summary
Supervised word sense disambiguation (WSD) performs well but depends on costly manually sense-tagged examples, which must be produced separately for every ambiguous word. Ng, Wang and Chan test whether word-aligned parallel text can supply this training data instead. The idea is that the different translations of an English word act as its sense labels. They use six English-Chinese corpora from the Linguistic Data Consortium (about 280 MB in total) and align words with GIZA++. For each noun, WordNet 1.7 senses that share a Chinese translation are merged, and English contexts aligned to the chosen translations become training examples for a naïve Bayes classifier. The method is evaluated on the nouns of the SENSEVAL-2 English lexical sample task and compared against training on the official hand-tagged data. On the hardest nouns, the gap with manual data is 14.0 points under a fair setup. It shrinks to 6.5 points when both sources have equal sense coverage. Much of the remaining gap comes from domain overlap between the official training and test data. This makes parallel text a credible, cheap alternative to manual sense tagging.

## Key points
- **Data:** Six LDC English-Chinese corpora, roughly 31.7M English words in total: three Hong Kong collections (news, laws and Hansards), plus Xinhua News, Sinorama and the English translation of Chinese Treebank.
- **Alignment cost:** Full alignment did not fit in memory. Hansards and Xinhua were therefore filtered to sentences containing target nouns, and the single GIZA++ run took about 40 hours on a 2.4 GHz machine with 2 GB RAM.
- **Sense lumping:** It was applied to all 29 SENSEVAL-2 nouns and cut the average sense count from 5.07 to 3.52. Seven nouns collapsed to a single translation and were excluded from WSD. Choosing translations took under 15 minutes per noun.
- **Classifier:** The WSD system of Lee and Ng (2002) used part-of-speech, surrounding words and local collocations as features, with naïve Bayes as the learner. Each training example was a 3-sentence English context.
- **Accuracy gaps on difficult nouns (manual minus parallel):**
  - 0.189 in the naive comparison (M1 vs P1).
  - 0.140 when training and test examples come from different documents (M2 vs P2).
  - 0.065 when manual data is limited to the per-sense counts available in parallel text (M3 vs P2).

## Interesting ideas
- **Translations as labels:** Treating target-language translations as sense labels gives a data-driven, application-grounded sense inventory. It could be reused to bootstrap labelled data for any language pair with parallel corpora, including Ukrainian-English.
- **Merging by translation:** Senses that translate identically are merged, so the task is reduced to distinctions that actually matter downstream, such as word choice in MT. This is a useful way to coarsen an overly fine lexicon.
- **Matching the sense distribution:** Parallel-text examples are sampled to follow the per-sense counts of the reference training set. This isolates the quality of the labels from the effect of class priors, a control worth copying whenever two data sources are compared.
- **Stepwise controlled comparisons:** The authors close the gap one factor at a time: 1. Remove same-document overlap. 2. Equalise sense coverage. 3. Inspect leftover domain effects, such as the collocation "Channel Tunnel" for *channel*. This turns a single headline number into an explanation of where the difference comes from.
- **Filtering before alignment:** When resources are limited, the authors keep only the sentences containing the target words before running alignment. This is a cheap trick to make large-corpus alignment feasible for focused lexical tasks.

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

Fields: [Natural language processing](https://md.research.elamur.ai/free/en/shelves/nlp/)

Topics: [🔀 Word sense disambiguation](https://md.research.elamur.ai/free/en/topics/word-sense-disambiguation/)

## Institutions

- [National University of Singapore](https://md.research.elamur.ai/free/en/universities/national-university-of-singapore/) — University, Singapore, Singapore · 61 works · 28,316 citations

## Cite

Cite as: Hwee Tou Ng, Bin Wang and Yee Seng Chan. 2003. Exploiting Parallel Texts for Word Sense Disambiguation: An Empirical Study. Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics.

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

```bibtex
@inproceedings{ng2003exploiting,
  title = {{Exploiting Parallel Texts for Word Sense Disambiguation: An Empirical Study}},
  author = {Hwee Tou Ng, Bin Wang and Yee Seng Chan},
  booktitle = {Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics},
  year = {2003},
  doi = {10.3115/1075096.1075154},
  url = {https://doi.org/10.3115/1075096.1075154}
}
```

- [← previous Loosely Tree-Based Alignment for Machine Translation](https://md.research.elamur.ai/free/en/papers/gildea-2003-loosely-tree-based-alignment-machine/)
- [next → A General Framework for Distributional Similarity](https://md.research.elamur.ai/free/en/papers/weeds-2003-general-framework-distributional-similarity/)
- [↑ back to papers of 2003](https://md.research.elamur.ai/free/en/papers/2003/)

## References

18 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Peter F. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, and Robert L. Mercer. 1991. Word-sense disambiguation using statistical methods. In Proceedings of the 29th Annual Meeting of the Association for Computational Linguistics, pages 264-270. [our card](https://md.research.elamur.ai/free/en/papers/brown-1991-word-sense-disambiguation-statistical-methods/)
2. Irina Chugur, Julio Gonzalo, and Felisa Verdejo. 2002. Polysemy and sense proximity in the Senseval-2 test suite. In Proceedings of the ACL SIGLEX Workshop on Word Sense Disambiguation: Recent Successes and Future Directions, pages 32-39. [find](https://scholar.google.com/scholar?q=Irina+Chugur%2C+Julio+Gonzalo%2C+and+Felisa+Verdejo%2E+2002%2E+Polysemy+and+sense+proximity+in+the+Senseval%2D2+test+suite%2E+In+Proceedings+of+the+ACL+SIGLEX+Workshop+on+Word+Sense+Disambiguation%3A+Recent+Success)
3. Mona Diab and Philip Resnik. 2002. An unsupervised method for word sense tagging using parallel corpora. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 255-262. [find](https://scholar.google.com/scholar?q=Mona+Diab+and+Philip+Resnik%2E+2002%2E+An+unsupervised+method+for+word+sense+tagging+using+parallel+corpora%2E+In+Proceedings+of+the+40th+Annual+Meeting+of+the+Association+for+Computational+Linguistics%2C+pag)
4. Philip Edmonds and Scott Cotton. 2001. SENSEVAL-2: Overview. In Proceedings of the Second International Workshop on Evaluating Word Sense Disambiguation Systems (SENSEVAL-2), pages 1-5. [find](https://scholar.google.com/scholar?q=Philip+Edmonds+and+Scott+Cotton%2E+2001%2E+SENSEVAL%2D2%3A+Overview%2E+In+Proceedings+of+the+Second+International+Workshop+on+Evaluating+Word+Sense+Disambiguation+Systems+%28SENSEVAL%2D2%29%2C+pages+1%2D5%2E)
5. Gerard Escudero, Lluis Marquez, and German Rigau. 2000. An empirical study of the domain dependence of supervised word sense disambiguation systems. In Proceedings of the Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora, pages 172-180. [find](https://scholar.google.com/scholar?q=Gerard+Escudero%2C+Lluis+Marquez%2C+and+German+Rigau%2E+2000%2E+An+empirical+study+of+the+domain+dependence+of+supervised+word+sense+disambiguation+systems%2E+In+Proceedings+of+the+Joint+SIGDAT+Conference+on+Em)
6. Radu Florian and David Yarowsky. 2002. Modeling consensus: Classifier combination for word sense disambiguation. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing, pages 25-32. [find](https://scholar.google.com/scholar?q=Radu+Florian+and+David+Yarowsky%2E+2002%2E+Modeling+consensus%3A+Classifier+combination+for+word+sense+disambiguation%2E+In+Proceedings+of+the+2002+Conference+on+Empirical+Methods+in+Natural+Language+Processi)
7. Nancy Ide, Tomaz Erjavec, and Dan Tufis. 2002. Sense discrimination with parallel corpora. In Proceedings of the ACL SIGLEX Workshop on Word Sense Disambiguation: Recent Successes and Future Directions, pages 54-60. [find](https://scholar.google.com/scholar?q=Nancy+Ide%2C+Tomaz+Erjavec%2C+and+Dan+Tufis%2E+2002%2E+Sense+discrimination+with+parallel+corpora%2E+In+Proceedings+of+the+ACL+SIGLEX+Workshop+on+Word+Sense+Disambiguation%3A+Recent+Successes+and+Future+Direction)
8. Adam Kilgarriff. 2001. English lexical sample task description. In Proceedings of the Second International Workshop on Evaluating Word Sense Disambiguation Systems (SENSEVAL-2), pages 17-20. [find](https://scholar.google.com/scholar?q=Adam+Kilgarriff%2E+2001%2E+English+lexical+sample+task+description%2E+In+Proceedings+of+the+Second+International+Workshop+on+Evaluating+Word+Sense+Disambiguation+Systems+%28SENSEVAL%2D2%29%2C+pages+17%2D20%2E)
9. Yoong Keok Lee and Hwee Tou Ng. 2002. An empirical evaluation of knowledge sources and learning algorithms for word sense disambiguation. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing, pages 41-48. [our card](https://md.research.elamur.ai/free/en/papers/lee-2002-empirical-evaluation-knowledge-sources-learning/)
10. Cong Li and Hang Li. 2002. Word translation disambiguation using bilingual bootstrapping. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 343-351. [find](https://scholar.google.com/scholar?q=Cong+Li+and+Hang+Li%2E+2002%2E+Word+translation+disambiguation+using+bilingual+bootstrapping%2E+In+Proceedings+of+the+40th+Annual+Meeting+of+the+Association+for+Computational+Linguistics%2C+pages+343%2D351%2E)
11. I. Dan Melamed. 2001. Empirical Methods for Exploiting Parallel Texts. MIT Press, Cambridge. [find](https://scholar.google.com/scholar?q=I%2E+Dan+Melamed%2E+2001%2E+Empirical+Methods+for+Exploiting+Parallel+Texts%2E+MIT+Press%2C+Cambridge%2E)
12. Rada F. Mihalcea and Dan I. Moldovan. 2001. Pattern learning and active feature selection for word sense disambiguation. In Proceedings of the Second International Workshop on Evaluating Word Sense Disambiguation Systems (SENSEVAL-2), pages 127-130. [find](https://scholar.google.com/scholar?q=Rada+F%2E+Mihalcea+and+Dan+I%2E+Moldovan%2E+2001%2E+Pattern+learning+and+active+feature+selection+for+word+sense+disambiguation%2E+In+Proceedings+of+the+Second+International+Workshop+on+Evaluating+Word+Sense+Di)
13. George A. Miller. (Ed.) 1990. WordNet: An on-line lexical database. International Journal of Lexicography, 3(4):235-312. [find](https://scholar.google.com/scholar?q=George+A%2E+Miller%2E+%28Ed%2E%29+1990%2E+WordNet%3A+An+on%2Dline+lexical+database%2E+International+Journal+of+Lexicography%2C+3%284%29%3A235%2D312%2E)
14. Franz Josef Och and Hermann Ney. 2000. Improved statistical alignment models. In Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, pages 440-447. [our card](https://md.research.elamur.ai/free/en/papers/och-2000-improved-statistical-alignment-models/)
15. Philip Resnik. 1999. Mining the Web for bilingual text. In Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics, pages 527-534. [find](https://scholar.google.com/scholar?q=Philip+Resnik%2E+1999%2E+Mining+the+Web+for+bilingual+text%2E+In+Proceedings+of+the+37th+Annual+Meeting+of+the+Association+for+Computational+Linguistics%2C+pages+527%2D534%2E)
16. Philip Resnik and David Yarowsky. 1997. A perspective on word sense disambiguation methods and their evaluation. In Proceedings of the ACL SIGLEX Workshop on Tagging Text with Lexical Semantics: Why, What, and How?, pages 79-86. [our card](https://md.research.elamur.ai/free/en/papers/resnik-1997-perspective-word-sense-disambiguation-methods/)
17. Philip Resnik and David Yarowsky. 2000. Distinguishing systems and distinguishing senses: New evaluation methods for word sense disambiguation. Natural Language Engineering, 5(2):113-133. [find](https://scholar.google.com/scholar?q=Philip+Resnik+and+David+Yarowsky%2E+2000%2E+Distinguishing+systems+and+distinguishing+senses%3A+New+evaluation+methods+for+word+sense+disambiguation%2E+Natural+Language+Engineering%2C+5%282%29%3A113%2D133%2E)
18. David Yarowsky, Silviu Cucerzan, Radu Florian, Charles Schafer, and Richard Wicentowski. 2001. The Johns Hopkins SENSEVAL2 system descriptions. In Proceedings of the Second International Workshop on Evaluating Word Sense Disambiguation Systems (SENSEVAL-2), pages 163-166. [find](https://scholar.google.com/scholar?q=David+Yarowsky%2C+Silviu+Cucerzan%2C+Radu+Florian%2C+Charles+Schafer%2C+and+Richard+Wicentowski%2E+2001%2E+The+Johns+Hopkins+SENSEVAL2+system+descriptions%2E+In+Proceedings+of+the+Second+International+Workshop+on+E)