---
title: "Automatic Verification and Augmentation of Multilingual Lexicons"
description: "The authors propose a fully automatic way to audit such lexicons and to suggest new entries. 16 references."
authors: ["Maryam Aminian", "Mohamed Al-Badrashiny", "Mona Diab"]
year: 2016
venue: "Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3)"
source: "https://aclanthology.org/W16-4810/"
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/aminian-2016-automatic-verification-augmentation-multilingual-lexicons/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/aminian-2016-automatic-verification-augmentation-multilingual-lexicons/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2016](https://md.research.elamur.ai/free/en/papers/2016/page/5/) / Automatic Verification and Augmentation of Multilingual Lexicons

# Automatic Verification and Augmentation of Multilingual Lexicons

**Authors:** Maryam Aminian, [Mohamed Al-Badrashiny](https://md.research.elamur.ai/free/en/authors/mohamed-al-badrashiny/) and [Mona Diab](https://md.research.elamur.ai/free/en/authors/mona-diab/) · **Year:** 2016 · **Venue:** [Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3)](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 2016](https://md.research.elamur.ai/free/en/venues/coling/2016/)) · **Publisher:** [International Committee on Computational Linguistics (ICCL)](https://md.research.elamur.ai/free/en/publishers/international-committee-on-computational-linguistics-iccl/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/W16-4810/) · [PDF](https://aclanthology.org/W16-4810.pdf) · [OpenAlex](https://openalex.org/W2785943593)

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

## Summary
Large multilingual lexicons are usually built partly by hand and partly by automatic means, and checking their quality is normally done through costly manual review. The authors propose a fully automatic way to audit such lexicons and to suggest new entries. They induce word-level translation triples by pivoting through English across two parallel corpora (Egyptian Arabic–English and MSA–English). The triples are filtered with a part-of-speech compatibility mapping. They are then widened with synonyms from WordNet (after word sense disambiguation), from K-means clusters of word2vec embeddings, and from CCA-projected cross-lingual embeddings. Comparing the resulting set, called TransDict, against an existing lexicon sorts each lexicon row into one of three groups: fully matched, partially matched, or unmatched. These groups point to coverage, likely errors, and augmentation candidates. Without human input, the method reaches F1 of 71.71% against the Tharwa EGY–MSA–EN lexicon and 54.46% against the Arabic entries of BabelNet. Manual checks indicate that many apparent misses are actually valid synonyms missing from the gold lexicon. This makes the method practical for maintaining resources for dialects with little data.

## Key points
- Data: Bolt-ARZ v3+v4 provides 3.5M EGY words of EGY–EN parallel text, and GALE phase4 provides about 60M MSA words of MSA–EN text. Monolingual corpora are EGY (about 260M words) and MSA (833M words), with English Gigaword 5th Edition used for the CCA model.
- Pipeline: MADAMIRA v1.0 lemmatizes and diacritizes the Arabic side, and TreeTagger handles English. GIZA++ produces the alignments, which are reduced to type level. Skip-gram word2vec vectors (300 dimensions, window 8) are clustered with K-means into either 500 or 9228 clusters.
- Main scores: F1 is 71.71% on gold Tharwa and 54.46% on BabelNet Arabic.
- Cluster granularity: 9228 clusters produce purer candidate lists than 500 clusters. Actual synonyms rise from 6.6% to 21.6% of EGY candidates per row. However, F1 falls with the finer setting because many synonyms are lost.
- Manual audit:
  - In 78.1% of sampled "non-matched" rows, at least one valid synonym was present. This puts estimated true matching accuracy at about 93.8%.
  - An expert judged 85% of completely unseen triples to be correct.

## Interesting ideas
- Build translation triples by going through a shared pivot language, and allow only specific cross-category POS mappings (e.g., adjective↔noun) rather than demanding identical tags; this keeps genuine divergences while removing most alignment noise.
- Sort every resource entry into fully matched, partially matched, or unmatched relative to independently induced evidence; the partial group is where errors and missing senses tend to surface.
- Treat gold incompleteness as part of the evaluation by having an expert review a sample of the system's apparent errors, then use that review to correct the accuracy estimate.
- Run sense disambiguation before expanding a word with its WordNet synonyms, so the expansion follows the intended synset instead of every synset.
- Choose cluster granularity as an explicit trade-off: finer clusters are purer but split the semantic space and reduce recall.

## For Mova

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

- **I1** (world): Cross-check compiled knowledge by pivoting between two resources that share a key, such as WordNet↔FrameNet through a common lemma, and accept only whitelisted category correspondences.
- **I2** (learn): Apply the same three-way triage to silver trees from the witness parsers, sending partially agreeing arcs to repair-rule induction or the LLM teacher instead of discarding them.
- **I3** (learn): When the evaluation gold is a lexicon or answer key, manually sample a portion of the "false negatives" to estimate how many are actually right, so scores are not underestimated.
- **I4** (frames): Expand a predicate to synonyms for VerbNet class lookup only after choosing its sense, which avoids role frames from unrelated meanings.
- **I5** (world): When grouping values or entities into classes for the knowledge graph, tune granularity against both purity and recall instead of assuming that finer is always better.

*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

- [George Washington University](https://md.research.elamur.ai/free/en/universities/george-washington-university/) — University, Washington, United States · 21 works · 7,106 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (2): [🌍 Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · 467 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 1,484 works

## Cite

Cite as: Maryam Aminian, Mohamed Al-Badrashiny and Mona Diab. 2016. Automatic Verification and Augmentation of Multilingual Lexicons. Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3).

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

```bibtex
@inproceedings{aminian2016automatic,
  title = {{Automatic Verification and Augmentation of Multilingual Lexicons}},
  author = {Maryam Aminian, Mohamed Al-Badrashiny and Mona Diab},
  booktitle = {Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial3)},
  year = {2016},
  url = {https://aclanthology.org/W16-4810/}
}
```

- [← previous Antoni Tronina, Mariusz Szmajdziński, Wprowadzenie do języka syryjskiego (Introduction to Syriac), Studia Biblica 6, Kielce 2003, Instytut Teologii Biblijnej Verbum, 153 pp., ISBN 83-915855-9-X](https://md.research.elamur.ai/free/en/papers/polanski-2026-antoni-tronina-mariusz-szmajdzinski-wprowadzenie/)
- [next → Creation of comparable corpora for English-Urdu, Arabic, Persian](https://md.research.elamur.ai/free/en/papers/abouammoh-2016-creation-comparable-corpora-english-urdu/)
- [↑ back to papers of 2016](https://md.research.elamur.ai/free/en/papers/2016/page/5/)

## References

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

1. William Black, Sabri Elkateb, Horacio Rodriguez, Musa Alkhalifa, Piek Vossen, Adam Pease, and Christiane Fellbaum. 2006. Introducing the arabic wordnet project. In Proceedings of the third international WordNet conference, pages 295–300. Citeseer. [find](https://scholar.google.com/scholar?q=William+Black%2C+Sabri+Elkateb%2C+Horacio+Rodriguez%2C+Musa+Alkhalifa%2C+Piek+Vossen%2C+Adam+Pease%2C+and+Christiane+Fellbaum%2E+2006%2E+Introducing+the+arabic+wordnet+project%2E+In+Proceedings+of+the+third+internation)
2. G. De Melo and G. Weikum. 2009. Towards a universal wordnet by learning from combined evidence. In Proceedings of the 18th ACM conference on Information and knowledge management, pages 513–522. ACM. [find](https://scholar.google.com/scholar?q=G%2E+De+Melo+and+G%2E+Weikum%2E+2009%2E+Towards+a+universal+wordnet+by+learning+from+combined+evidence%2E+In+Proceedings+of+the+18th+ACM+conference+on+Information+and+knowledge+management%2C+pages+513%E2%80%93522%2E+ACM%2E)
3. Mona Diab, Mohamed Al-Badrashiny, Maryam Aminian, Mohammed Attia, Pradeep Dasigi, Heba Elfardy, Ramy Eskander, Nizar Habash, Abdelati Hawwari, and Wael Salloum. 2014. Tharwa: A Large Scale Dialectal Arabic-Standard Arabic-English Lexicon. In LREC. [find](https://scholar.google.com/scholar?q=Mona+Diab%2C+Mohamed+Al%2DBadrashiny%2C+Maryam+Aminian%2C+Mohammed+Attia%2C+Pradeep+Dasigi%2C+Heba+Elfardy%2C+Ramy+Eskander%2C+Nizar+Habash%2C+Abdelati+Hawwari%2C+and+Wael+Salloum%2E+2014%2E+Tharwa%3A+A+Large+Scale+Dialectal+A)
4. Manaal Faruqui and Chris Dyer. 2014. [Improving vector space word representations using multilingual correlation](https://md.research.elamur.ai/free/en/papers/faruqui-2014-improving-vector-space-word-representations/). In Proceedings of EACL.
5. Christiane Fellbaum. 1998. WordNet. Wiley Online Library. [find](https://scholar.google.com/scholar?q=Christiane+Fellbaum%2E+1998%2E+WordNet%2E+Wiley+Online+Library%2E)
6. 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/). arXiv preprint arXiv:1301.3781. [↗](https://arxiv.org/abs/1301.3781)
7. Roberto Navigli and Simone Paolo Ponzetto. 2012. Babelnet: The automatic construction, evaluation and application of a wide-coverage multilingual semantic network. Artificial Intelligence, 193:217–250. [↗](https://doi.org/10.1016/j.artint.2012.07.001)
8. Franz Josef Och and Hermann Ney. 2000. [Improved statistical alignment models](https://md.research.elamur.ai/free/en/papers/och-2000-improved-statistical-alignment-models/). In Proceedings of the 38th Annual Meeting on Association for Computational Linguistics, pages 440–447. Association for Computational Linguistics.
9. Robert Parker, David Graff, Junbo Kong, Ke Chen, and Kazuaki Maeda. 2011. English gigaword fifth edition. [↗](https://doi.org/10.35111/wk4f-qt80)
10. Arfath Pasha, Mohamed Al-Badrashiny, Mona Diab, Ahmed El Kholy, Ramy Eskander, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan M. Roth. 2014. [MADAMIRA: A Fast, Comprehensive Tool for Morphological Analysis and Disambiguation of Arabic](https://md.research.elamur.ai/free/en/papers/pasha-2014-madamira-fast-comprehensive-tool-morphological/). In Proceedings of LREC, Reykjavik, Iceland.
11. Ted Pedersen, Satanjeev Banerjee, and Siddharth Patwardhan. 2005. Maximizing semantic relatedness to perform word sense disambiguation. University of Minnesota supercomputing institute research report UMSI, 25:2005. [find](https://scholar.google.com/scholar?q=Ted+Pedersen%2C+Satanjeev+Banerjee%2C+and+Siddharth+Patwardhan%2E+2005%2E+Maximizing+semantic+relatedness+to+perform+word+sense+disambiguation%2E+University+of+Minnesota+supercomputing+institute+research+report)
12. Benoı̂t Sagot and Darja Fišer. 2011a. Classification-based extension of wordnets from heterogeneous resources. In Human Language Technology Challenges for Computer Science and Linguistics, pages 396–407. Springer. [find](https://scholar.google.com/scholar?q=Beno%C4%B1%CC%82t+Sagot+and+Darja+Fis%CC%8Cer%2E+2011a%2E+Classification%2Dbased+extension+of+wordnets+from+heterogeneous+resources%2E+In+Human+Language+Technology+Challenges+for+Computer+Science+and+Linguistics%2C+pages+396%E2%80%93)
13. Benoı̂t Sagot and Darja Fišer. 2011b. Extending wordnets by learning from multiple resources. In LTC’11: 5th Language and Technology Conference. [find](https://scholar.google.com/scholar?q=Beno%C4%B1%CC%82t+Sagot+and+Darja+Fis%CC%8Cer%2E+2011b%2E+Extending+wordnets+by+learning+from+multiple+resources%2E+In+LTC%E2%80%9911%3A+5th+Language+and+Technology+Conference%2E)
14. Benoı̂t Sagot and Darja Fišer. 2012. Automatic extension of wolf. In GWC2012-6th International Global Wordnet Conference. [find](https://scholar.google.com/scholar?q=Beno%C4%B1%CC%82t+Sagot+and+Darja+Fis%CC%8Cer%2E+2012%2E+Automatic+extension+of+wolf%2E+In+GWC2012%2D6th+International+Global+Wordnet+Conference%2E)
15. I. Saleh and N. Habash. 2009. Automatic extraction of lemma-based bilingual dictionaries for morphologically rich languages. In Third Workshop on Computational Approaches to Arabic Script-based Languages at the MT Summit XII, Ottawa, Canada. [find](https://scholar.google.com/scholar?q=I%2E+Saleh+and+N%2E+Habash%2E+2009%2E+Automatic+extraction+of+lemma%2Dbased+bilingual+dictionaries+for+morphologically+rich+languages%2E+In+Third+Workshop+on+Computational+Approaches+to+Arabic+Script%2Dbased+Langua)
16. Helmut Schmid. 1995. Treetagger— a language independent part-of-speech tagger. Institut für Maschinelle Sprachverarbeitung, Universität Stuttgart, 43:28. [find](https://scholar.google.com/scholar?q=Helmut+Schmid%2E+1995%2E+Treetagger%E2%80%94+a+language+independent+part%2Dof%2Dspeech+tagger%2E+Institut+fu%CC%88r+Maschinelle+Sprachverarbeitung%2C+Universita%CC%88t+Stuttgart%2C+43%3A28%2E)

*Updated 24 October 2026*