Method · machine learning (optimal transport)

Gromov-Wasserstein alignment of written-word and Latin co-occurrence spaces

Problem Track A can't solve: whether there is a word-level map at all. Track A assumes signs inside words; this needs no key and no segmentation, only co-occurrence.

Machine learning (optimal transport)Version 1.0Licence MIT2 runsUsed on 1 mystery

Code Python

Download module (.py)gw_t43.pyView source

Runs and results 2 runs

Run	Mystery	Measured	Result	Against	Date
RUN0DCF8N67G74D	Voynich	Held-out word accuracy of the coupling's word map (bar 0.5)	word-spacing control 0.003 (GW 0.034); stream-spacing control 0.0014 (GW 0.037)	Gate 0.5
RUN0DCF9GEQF284	Voynich	Held-out word accuracy of the coupling's word map (bar 0.5)	word-spacing control 0.039; stream-spacing control 0.0007	Generation 1: 0.003

Applied to

Voynich Manuscript

3 datasetsHerbal A and Herbal B paragraph text: training and held-out pages, Latin genre texts, Naibbe cipher tables and example plaintexts (greshko/naibbe-cipher, commit 719f2f4c)

References

Alvarez-Melis, D., Jaakkola, T. 2018. Gromov-Wasserstein Alignment of Word Embedding Spaces. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.